Timeline of Grok

This is a timeline of Grok, an AI assistant developed by xAI. First launched in November 2023, it is powered by a large language model. The timeline traces the evolution of Grok from the 1961 origin of the term in Heinlein's Stranger in a Strange Land through xAI's founding and Grok's initial 2023 launch to its subsequent development through mid-2026. It records major model releases and technical milestones, xAI's corporate history including its acquisition of Twitter, Inc. and subsequent merger with SpaceX, comparative benchmarking against rival models from OpenAI, Google, and Anthropic, and Grok's expanding integration across platforms including X, Azure, Tesla, Inc. vehicles, and numerous third-party developer tools. The timeline also documents the substantial controversies, privacy and security incidents, and regulatory and legal actions that have accompanied Grok's growth — particularly around AI-generated non-consensual imagery — alongside its expanding role in multimodal, agentic, and reasoning-oriented artificial intelligence.

Sample questions

The following are some interesting questions that can be answered by reading this timeline:

• How has Grok's core model lineage evolved from launch to its latest coding-focused and multimodal variants?
  • Sort the full timeline by "Event type" and look for the group of rows with value "Release".
  • You will see the progression from Grok 1 (November 2023) through Grok 1.5, Grok-1.5V, Grok 2, Grok 3, Grok 4, Grok 4.1, grok-code-fast-1 (August 2025), Grok 4.3, and Grok 4.5, tracing growth in context length, multimodality, agentic tool use, and reasoning capability.
• What new capabilities has xAI integrated into Grok's product ecosystem over time?
  • Sort the full timeline by "Event type" and look for the group of rows with value "Integration".
  • You will see additions like PromptIDE (November 2023), the Grok API public beta (November 2024), Aurora image generation (December 2024), and the Grok Imagine API (January 2026).
• Across which platforms has Grok been deployed since its 2023 launch?
  • Sort the full timeline by "Event type" and look for the group of rows with value "Deployment".
  • You will see rollouts to X Premium+ subscribers (December 2023), the Grok app on Android (February 2025), Grok Imagine going free on X (August 2025), Grok 4 on Microsoft Azure AI Foundry (September 2025), and Grok's arrival in Tesla vehicles in the UK, Europe, and India (February–August 2026).
• What controversies has Grok been involved in, and how have they evolved over time?
  • Sort the full timeline by "Event type" and look for the group of rows with value "Controversy".
  • You will see the fabricated "Iran Strikes Tel Aviv" headline (April 2024), medical-image privacy concerns (November 2024), Grok 3 benchmark manipulation allegations (February 2025), antisemitic outputs (July 2025), and the mass generation of non-consensual sexualised images (January 2026).
• What regulatory actions have governments taken against Grok, and why?
  • Sort the full timeline by "Event type" and look for the group of rows with value "Regulation".
  • You will see Indonesia's ban (January 2026), Malaysia's ban (January 2026), Canada's Privacy Commissioner finding (June 2026), and joint UK/EU investigations into Grok's data handling and deepfake safeguards (February 2026).
• What legal action has been brought against or by xAI over Grok, and by whom?
  • Sort the full timeline by "Event type" and look for the group of rows with value "Litigation".
  • You will see a San Jose class-action suit by anonymous plaintiffs (January 2026), Baltimore's municipal consumer-protection suit (March 2026), xAI's own suit against a user for generating CSAM (July 2026), a UK MP's case against non-consensual deepfakes (July 2026), a ruling protecting the San Jose plaintiffs' anonymity (July 2026), and xAI's unsuccessful challenge to a Minnesota "nudification" ban (July 2026).
• Which senior leaders and co-founders have left xAI, and why?
  • Sort the full timeline by "Event type" and look for the group of rows with value "Personnel".
  • You will see a wave of departures following Grok's antisemitic-posting controversy in mid-2025 (X CEO Linda Yaccarino, CFO Mike Liberatore, general counsel Robert Keele), co-founder Igor Babuschkin's exit to launch an AI safety investment firm (August 2025), and the completion of a total co-founder exodus by March 2026 — all eleven of xAI's original founders, including Tony Wu, Jimmy Ba, and Christian Szegedy, having left within about a year of the company's founding.
• What background context and competing developments shaped Grok's emergence before its 2023 launch?
  • Sort the full timeline by "Event type" and look for the group of rows with value "Context".
  • You will see Elon Musk's endorsement of the Future of Life Institute's AI moratorium letter (March 2023) and his early "TruthGPT" announcement (April 2023).
• What has Elon Musk publicly announced about Grok's future development?
  • Sort the full timeline by "Event type" and look for the group of rows with value "Announcement".
  • You will see the Grok 5 training announcement (August 2025) and the Grok 4.2 performance projection (February 2026).
• How does Grok's own marketing compare to independent data on where it actually stood against rivals at key release moments?
  • Sort the full timeline by "Event type" and look for the group of rows with value "Comparison".
  • You will see Grok-1 entering a field still led by GPT-4 (November 2023), Grok 3's self-reported edge over o1 and Claude 3.5 Sonnet (February 2025), Grok 4's claimed lead over Claude Opus 4 and Gemini 2.5 Pro (July 2025), Grok 4.1's LMArena standing against Claude and GPT rivals (November 2025), and Grok 4.5's independently-run benchmark results against Claude Fable 5, Opus 4.7/4.8, and GPT-5.5 (July 2026).
• Other events are described under the following types: "Naming", "Founding", "OpenSource", "Commercialization", "Product launch", "Privacy", "Competition", "Security", "Restriction", "Commentary", "Adoption", "Acquisition".

Big picture

Time period	Development summary	More details
2023	Prelude and launch	In April, Elon Musk announces plans for TruthGPT, a “maximum truth-seeking AI” aimed at countering perceived political bias in systems like ChatGPT. Soon after, he founds X.AI Corp. in Nevada to pursue this vision. On November 3, xAI releases Grok-1, a large language model developed in Rust and Python, integrated with X (formerly Twitter). Distinguished by humor and personality, Grok represents Musk’s effort to combine reasoning and character within conversational AI. Days later, xAI introduces PromptIDE, an interpretability environment allowing developers to inspect model behavior through visual analytics, token tracking, and shared prompt libraries. Together, Grok-1 and PromptIDE establish xAI’s foundation—merging technical transparency with a distinctive, personality-driven approach that set Grok apart from competitors like OpenAI’s ChatGPT and Anthropic’s Claude.
2024	Expansion, multimodality, and scrutiny	xAI expands Grok’s capabilities and reach. The company open-sources the 314-billion-parameter Grok-1 model in March, followed weeks later by Grok 1.5, featuring a 128K-token context and stronger reasoning. Grok 1.5V adds multimodal understanding, enabling text–image analysis across documents, diagrams, and photos. However, Grok’s growing influence also brings scrutiny—its generation of a false “Iran Strikes Tel Aviv” headline exposes the risks of unmoderated AI-generated news. Later, Grok 2 and 2 Mini deepen multimodality, integrating Black Forest Labs’ FLUX for text-to-image synthesis. November’s Grok API beta opens programmatic access to developers, while Aurora, an autoregressive image model, extends creative applications on X. Despite innovations, Musk’s call for users to upload medical images trigger privacy alarms, underscoring tensions between experimentation and ethical safeguards.
2025	Commercialization and controversy	Grok matures into a global AI ecosystem. The Grok 3 release in February, trained on xAI’s Colossus supercomputer using over 100,000 Nvidia GPUs, introduces “Think Mode” and real-time “DeepSearch,” surpassing GPT-4o in reasoning benchmarks. Yet, allegations of benchmark manipulation, antisemitic outputs, and privacy leaks tarnish its image. xAI expands commercially, launching xAI for Government with a $200 million U.S. Defense contract and deploying Grok 4 on Microsoft Azure AI Foundry for enterprise use. Meanwhile, Grok Imagine democratizes text-to-video generation but faces security issues after cybercriminals exploit vulnerabilities through prompt injection. Despite recurring controversies, Grok consolidates its role as a high-performance, multimodal AI suite—symbolizing both the technological ambition and ethical turbulence of the post-ChatGPT era. In September, Grok-5 begins training.
2026	Acquisition, regulatory reckoning, and ecosystem maturity	xAI closes a $20 billion Series E round in January before being acquired by SpaceX in February, merging Grok with SpaceX's rockets, Starlink, and the X platform under a long-term vision of space-based AI compute. The year brings intense regulatory and legal fallout from Grok's image-generation tools: Indonesia and Malaysia block Grok outright, the UK and EU open investigations, Canada's privacy regulator finds xAI violated federal law, and a wave of litigation unfolds, including a San Jose class action, xAI's own suit against a user for CSAM, a UK MP's case against non-consensual deepfakes, and xAI's fight against a Minnesota "nudification" ban. Amid this, xAI deepens its Pentagon relationship with a classified-systems deal and expands Grok across Tesla vehicles, Microsoft 365 and Google Workspace, and a growing roster of third-party platforms (eToro, Interactive Brokers, Databricks, Gopuff, Vapi, Amazon Bedrock). The company ships Grok 4.3 and Grok 4.5, open-sources Grok Build, and rolls out Skills, Connectors, Automations, and Build Mode, positioning Grok as a full agentic ecosystem even as its safety controversies intensify.

Full timeline

Inclusion criteria

This timeline focuses on events materially shaping Grok's technical development, product ecosystem, corporate structure, and public reception, rather than attempting to be an exhaustive record of every xAI announcement or Grok-related news item. The following criteria guide what is included and excluded.

Included:

• Model releases and technical milestones — every named Grok model release (Grok-1 through Grok 4.5 and their variants, e.g. mini, Fast, Heavy) is included, along with major architectural or capability changes (context window expansions, multimodality, tool use, agentic capabilities) reported in the release's own launch materials or independent coverage.
• Product and platform launches — new consumer- or developer-facing products (Grok Imagine, Grok Voice, Grok Build, Skills, Connectors, Automations, Build Mode, and the various API offerings) are included at their initial launch and at major subsequent milestones (e.g. going free, reaching general availability).
• Third-party integrations — partnerships that bring Grok into external platforms (coding tools, messaging apps, financial platforms, cloud providers) are included when they represent a first-of-its-kind integration or a partnership of substantial scale (e.g. millions of users, a major enterprise platform).
• Funding and corporate transactions — funding rounds, acquisitions, and major valuation events (Series B/C/E, the X acquisition, the SpaceX merger, the planned IPO) are included comprehensively given their centrality to the story of xAI's growth.
• Controversies, security incidents, and privacy issues — included when they receive independent (non-xAI) coverage and represent a meaningful lapse, harm, or public reaction, rather than routine bug reports.
• Regulatory and legal actions — bans, investigations, rulings, and lawsuits (whether brought against xAI or by xAI) are included comprehensively, given how central this friction has become to Grok's development since late 2025; even litigation still in early stages (e.g. newly filed complaints) is included given its relevance to the timeline's throughline on safety and accountability.
• Personnel changes — departures and appointments of co-founders, C-suite executives, and other senior leadership are included, especially where they form or contribute to a documented pattern (e.g. the full co-founder exodus, the post-antisemitism-controversy departure wave).
• Comparative benchmarking — dedicated Comparison rows are included at key model-release inflection points to independently or self-reportedly situate Grok against contemporaneous rivals (GPT, Gemini, Claude), rather than embedding comparative claims piecemeal across unrelated rows.
• Context and prelude events — events preceding xAI's founding (the coining of "grok," early AI-moratorium debates, competing lab announcements) are included when they directly motivate or foreshadow Grok's creation and development.

Not included:

• Routine or incremental product updates — minor UI tweaks, small pricing adjustments, or small feature refinements that don't materially change Grok's capabilities or reach are omitted unless independently notable.
• Every employee hire or departure — only co-founder, C-suite, and other senior leadership personnel changes are tracked; departures of individual engineers or non-senior staff are excluded unless part of an already-documented broader pattern.
• Duplicate or near-duplicate integration announcements — when multiple similar third-party integrations occur in quick succession without independent significance, only the most notable or representative examples are kept, to avoid a long undifferentiated list.
• Single-source or unverified claims — rumors, unconfirmed reports, or claims resting on a single low-quality source (e.g. content-marketing sites, SEO explainers) are excluded until corroborated by a stronger source; see the wiki's general source-quality guidance.
• Routine financial or usage statistics — recurring metrics (e.g. monthly active user counts, quarterly revenue figures) are not tracked as standalone rows unless the specific data point is unusual, triggers independent discussion, or the time point itself is significant.
• General AI industry news not specific to Grok or xAI — broader developments at other labs are only included where they materially affected Grok's competitive position, prompted a direct response, or are needed as backdrop context (e.g. the DeepSeek-R1 release ahead of Grok 3, GPT-5's benchmark comparison to Grok 4).
• Wikipedia-sourced claims — per wiki-wide policy, Wikipedia itself is never used as a citation source for any row.
• Speculative or unannounced future events — plans, rumors, or leaks about unreleased products are excluded unless xAI or Musk has made a public, citable statement about them (in which case they may appear as an Announcement-type row, framed as a stated intention rather than a confirmed fact).

Year	Event type	Details
1961	Naming	Robert A. Heinlein publishes the science fiction novel Stranger in a Strange Land, coining the verb "to grok" — meaning literally "to drink" but understood more broadly as "to understand profoundly and intuitively" — to describe an intuitive comprehension that transcends ordinary human understanding. The word is later incorporated into English-language dictionaries and enters popular and technological vocabulary. Decades later, the term would inspire Elon Musk's 2023 AI chatbot launch, reflecting the same theme of deep understanding between humans and artificial intelligence.
2023 (March 9)	Founding	xAI is founded with a mission to accelerate human scientific discovery and advance collective understanding of the universe. The company's early internal project was codenamed "TruthGPT."
2023 (March 22)	Context	Elon Musk endorses an open letter from the Future of Life Institute calling for a six-month moratorium on training AI frontier models exceeding the capabilities of OpenAI's GPT-4. Signed by over 33,000 people, including researchers and tech leaders, it warns that unchecked AI development poses serious societal and existential risks. The letter urges creating shared safety protocols, independent oversight, and strong AI governance before continuing the AI race. It advocates focusing on transparency, reliability, and alignment rather than unregulated model scaling.
2023 (April 17)	Context	Elon Musk announces plans to create “TruthGPT,” an AI chatbot designed to counter perceived political bias in existing systems like ChatGPT. In a Fox News interview with Tucker Carlson, Musk describes it as a “maximum truth-seeking AI” aimed at understanding the universe, arguing that such an AI would be less likely to harm humanity. He reiterates his long-standing concerns about AI’s dangers and advocates for regulation. Around the same time, Musk founds X.AI Corp. in Nevada to pursue his new AI ventures.
2023 (November 3)	Release	xAI announces Grok, an AI chatbot modeled on The Hitchhiker's Guide to the Galaxy, designed with wit and a "rebellious streak," offering real-time knowledge via the X platform and willingness to answer questions other AI systems reject. Grok is powered by Grok-1, developed over four months and preceded by an internal 33-billion-parameter prototype, Grok-0. Grok-1 surpasses ChatGPT-3.5 and Inflection-1 on benchmarks including HumanEval (63.2%) and MMLU (73%), trailing only larger models like GPT-4. On a hand-graded 2023 Hungarian national high school math exam withheld from training data, Grok-1 scores 59%, ahead of Claude 2's 55% but behind GPT-4's 68%. Built on a custom Kubernetes, Rust, and JAX training stack, Grok launches in early access to a limited number of U.S. users.
2023 (November 6)	Integration	xAI announces PromptIDE, an integrated development environment for prompt engineering and interpretability research. Designed to work with Grok-1, it provides engineers and researchers with transparent access to model behavior. The IDE includes a Python editor, an SDK for advanced prompting, and visual analytics showing tokenization, sampling probabilities, and attention masks. It also features automatic saving, versioning, and data upload support. Users can share prompts publicly to foster community collaboration. PromptIDE is initially released to members of xAI's early access program.
2023 (November)	Comparison	Stanford University's AI Index Report 2024 offers independent, contemporaneous context for Grok-1's benchmark claims at launch. On the HELM (Holistic Evaluation of Language Models) leaderboard as of January 2024, GPT-4 (0613) leads with a mean win rate of 0.96, well ahead of Anthropic's Claude 2.0 (0.68) and Claude v1.3 (0.72) — models Grok-1's own launch claims did not directly address. On MMLU, GPT-4 held the state-of-the-art score for most of 2023 before being surpassed by Google's Gemini Ultra (90.0%) in December 2023, a month after Grok-1's launch; Grok-1's self-reported MMLU score of 73% trails both. On the Chatbot Arena Leaderboard, GPT-4-Turbo ranks as the most preferred model among the general public as of early 2024, with Claude-2.1 and Gemini Pro also outperforming older GPT-3.5 and Claude-1 checkpoints. The report underscores that Grok-1 entered a field where GPT-4 held a substantial and largely uncontested performance lead across most standard benchmarks, with Claude 2 and Gemini as the more direct emerging challengers.
2023 (December 7)	Deployment	xAI rolls out Grok-1 to Premium+ subscribers on the social media platform X (formerly Twitter).
2024 (March 17)	OpenSource	xAI publicly releases the weights and architecture of Grok-1, its 314-billion-parameter Mixture-of-Experts LLM, under the Apache 2.0 open-source license. Trained from scratch using a custom JAX and Rust framework, Grok-1 represents xAI's foundational model for future Grok systems. The release includes the raw pre-training checkpoint, meaning the model is not fine-tuned for tasks such as conversation. Grok-1 activates about 25% of its parameters per token, balancing efficiency and scale. Developers can access setup instructions and implementation guidance through xAI's official GitHub repository at github.com/xai-org/grok.
2024 (March 28)	Release	xAI announces Grok-1.5, expanding the model's context window to 128,000 tokens — a 16x increase — while improving reasoning, coding, and math performance. The model scores 50.6% on the MATH benchmark, 90% on GSM8K, and 74.1% on HumanEval, notable gains over Grok-1, and achieves perfect recall in Needle-In-A-Haystack retrieval tests across the full context window. Grok-1.5 is built on a custom distributed training framework using JAX, Rust, and Kubernetes, with an orchestrator designed to automatically detect and eject failing nodes during large-scale training runs. The model rolls out first to early testers and existing Grok users on X.
2024 (March 29)	Announcement	Elon Musk states via X that Grok 2, the successor to the yet-unreleased Grok-1.5, is already in training and expected to "exceed current AI on all metrics." The announcement comes as Grok-1.5 itself is set to roll out to early testers and existing Grok users on X in the coming days.
2024 (April 4)	Controversy	Grok is found to generate a false headline claiming that “Iran Strikes Tel Aviv with Heavy Missiles,” which is then promoted on X’s official trending news section. The event occurs shortly after Israel’s airstrike on Iran’s embassy in Syria, making the fake headline appear plausible. However, no such attack happens. The incident highlights the risks of removing human editors—a policy change implemented by Elon Musk after acquiring Twitter. Previously, Twitter had relied on human curation and partnerships with Associated Press and Reuters to verify trending news content.
2024 (April 12)	Release	xAI launches Grok-1.5V, its first-generation multimodal model that combines advanced text understanding with visual processing. Grok-1.5V can interpret documents, diagrams, charts, screenshots, and photographs, bridging digital and physical information. It demonstrates competitive performance across multiple domains, including multi-disciplinary reasoning and real-world spatial understanding, excelling in xAI's RealWorldQA benchmark. Evaluations are conducted in a zero-shot setting without chain-of-thought prompting.
2024 (May 5)	Deployment	Grok-1.5 becomes available on X.
2024 (May 26)	Commercialization	xAI closes a $6 billion Series B funding round, with participation from Valor Equity Partners, Vy Capital, Andreessen Horowitz, Sequoia Capital, Fidelity Investments, Alwaleed bin Talal, and Kingdom Holding Company, among others. xAI states the funds will go toward bringing its first products to market, building infrastructure, and accelerating research and development.
2024 (August 13)	Release	xAI releases an early preview of Grok-2 and a smaller Grok-2 mini, a significant step up from Grok-1.5 in chat, coding, and reasoning. An early version tested on the LMSYS Chatbot Arena leaderboard under the codename "sus-column-r" outperforms both Claude 3.5 Sonnet and GPT-4 Turbo on overall Elo score. Grok-2 also shows state-of-the-art performance among the compared models on visual math reasoning (MathVista) and document-based question answering (DocVQA). Both models roll out in beta to X Premium and Premium+ subscribers via a redesigned Grok interface, with enterprise API access to follow later in the month; xAI states it is experimenting with Black Forest Labs' FLUX.1 model to expand Grok's image capabilities on X.
2024 (November 4)	Integration	xAI launches the public beta of its Grok API, allowing developers to access Grok foundation models programmatically. During the beta, running until the end of 2024, all users receive $25 of free API credits per month, with existing prepaid credit holders receiving the equivalent added on top. The preview includes grok-beta, a text model with a 128,000-token context length, function calling, and system prompt support, with a multimodal vision-capable version following the next week. The API is REST-compatible with OpenAI and Anthropic, simplifying migration by only requiring a change to the base URL and API key.
2024 (November 4)	Controversy	Elon Musk urges users to upload medical images to Grok AI for diagnostic analysis, sparking privacy and accuracy concerns. Physicians testing the tool find mixed results—some note impressive reasoning with detailed prompts, while others report incorrect diagnoses. Experts warn that sharing medical data on X could violate privacy regulations like Health Insurance Portability and Accountability Act and General Data Protection Regulation. Grok's healthcare potential draws both interest and skepticism, with Musk suggesting future integration with Neuralink. Regulators in Europe and U.S. states also question xAI's data practices and misinformation controls.
2024 (December 9)	Integration	xAI releases Aurora, an autoregressive image generation model that enhances Grok's capabilities on the X platform. Aurora is a mixture-of-experts network trained on billions of text–image pairs from the internet to predict the next token in interleaved data. This training enables it to produce highly photorealistic images and accurately follow textual instructions. The model also supports multimodal inputs, allowing users to draw inspiration from or directly edit their own images.
2024 (December 12)	Release	xAI rolls out an improved version of Grok-2 to all users on X for free, three times faster than the prior version with better accuracy, instruction-following, and multilingual support. The company also launches a new "Grok analysis" button appearing on posts across the X timeline, letting users request contextual explanations of trending content directly. Via the API, xAI introduces `grok-2-1212` and `grok-2-vision-1212`, improved versions of its enterprise models, while lowering pricing to $2 per million input tokens and $10 per million output tokens. Aurora image generation is announced as coming soon to the API.
2024 (December 23)	Commercialization	xAI closes a $6 billion Series C funding round, with participation from Andreessen Horowitz, BlackRock, Fidelity Management & Research Company, Kingdom Holdings, Lightspeed Venture Partners, MGX (company), Morgan Stanley, Oman Investment Authority, Qatar Investment Authority, Sequoia Capital, Valor Equity Partners, and Vy Capital, alongside strategic investors Nvidia and AMD. xAI reports that its Colossus supercomputer, built using 100,000 Nvidia Hopper GPUs, became fully operational in 122 days and began running workloads just 19 days after its first servers were delivered, with plans to double it to 200,000 GPUs using Nvidia's Spectrum-X networking platform. The company states its next model, Grok 3, is in training.
2025 (January 27)	Context	Chinese AI lab DeepSeek releases an open version of DeepSeek-R1, a reasoning model it claims matches or exceeds OpenAI's o1 on benchmarks including AIME, MATH-500, and SWE-bench Verified, while being 90–95% cheaper to access via API. Built with 671 billion parameters and released under an MIT license via Hugging Face alongside smaller "distilled" versions as small as 1.5 billion parameters, R1 rapidly generates over 500 derivative models and 2.5 million combined downloads on the platform. The release, arriving days after the outgoing Joe Biden administration proposes stricter export rules on AI technology for Chinese firms, intensifies competitive pressure in the reasoning-model race that xAI's Grok 3 would enter three weeks later. An AI researcher notes the trend suggests Chinese labs will continue to be "fast followers," while R1 remains subject to Chinese regulatory content restrictions, declining to answer questions on topics such as Tiananmen Square or Taiwan's autonomy.
2025 (February 5)	Deployment	Grok app becomes available for pre-registration on the Google Play Store, one month after its iOS debut. Android users in select countries—including Australia, Canada, India, the Philippines, and Saudi Arabia—can access the limited beta, while U.S. users may pre-register to be notified upon release. The app offers the Grok 2 model, matching the free and paid tiers already available on X (formerly Twitter). The delayed Android release reflects xAI's continued iOS-first rollout strategy but marks progress toward broader platform availability.
2025 (February 17)	Release	xAI officially releases Grok 3, the third generation of its large language model. Built on the company's Colossus supercomputer using over 100,000 Nvidia Hopper GPUs, Grok 3 introduces advanced reasoning through a new Think mode and real-time DeepSearch capabilities for up-to-date internet information. It achieves industry-leading benchmark scores, including 79.4 on LiveCodeBench (coding) and 99.3 on AIME 2024 (math), outperforming GPT-4o and DeepSeek-R1. The model remains proprietary, marking xAI's most powerful and capable AI release to date.
2025 (February 19)	Release	xAI releases an early preview of Grok 3, alongside a smaller Grok 3 mini, trained on the Colossus supercluster with 10x the compute of prior state-of-the-art models. The reasoning-enabled "Think" variant achieves 93.3% on AIME 2025 and 84.6% on GPQA using high test-time compute, while the non-reasoning model tops several benchmarks among non-reasoning models and reaches a 1402 Elo score on Chatbot Arena. xAI also introduces DeepSearch, its first agent, designed to synthesize information across the web into concise research reports. Grok 3 launches with a 1-million-token context window, immediately available to X Premium and Premium+ users, with API access to follow.
2025 (February 19)	Comparison	xAI's own launch materials for Grok 3 provide the fullest picture of how it claimed to stack up against contemporaneous rivals. In non-reasoning mode, Grok 3 Beta reports 52.2% on AIME'24, 75.4% on GPQA, and 57.0% on LiveCodeBench, ahead of Anthropic's Claude 3.5 Sonnet (16.0%, 65.0%, 40.2% respectively), OpenAI's GPT-4o (9.3%, 53.6%, 32.3%), Google's Gemini 2.0, and DeepSeek-V3 across the same benchmarks. With reasoning enabled, Grok 3 (Think) claims 93.3% on both AIME'24 and AIME'25 at its highest test-time compute setting, ahead of OpenAI's o3-mini (high) at 87.3% and 86.5% respectively, o1 at 83.3% and 79%, and DeepSeek-R1 at 79.8% and 70%. On GPQA, Grok 3 (Think)'s 84.6% edges out o3-mini (high) at 79.7% and o1 at 78%. These figures come directly from xAI's own launch comparison charts rather than independent evaluation, and the specific AIME methodology behind them would draw public criticism days later for omitting the cons@64 context needed to assess whether the comparison was fair.
2025 (February 22)	Controversy	Controversy arises after an OpenAI employee accuses xAI of misleadingly presenting Grok 3's benchmark results. xAI had published a graph showing Grok 3 Reasoning Beta and Grok 3 mini Reasoning outperforming OpenAI's o3-mini-high on the AIME 2025 math benchmark. Critics note xAI omitted cons@64 scores, which allow multiple attempts and usually boost performance, making Grok 3's advantage appear overstated. xAI co-founder Igor Babuschkin defends the company, arguing OpenAI had done similar comparisons. Experts like Nathan Lambert (researcher) emphasize that benchmark results often lack context on computational costs and limitations.
2025 (March 28)	Acquisition	xAI announces it has acquired Twitter, Inc. in an all-stock transaction, with the combination valuing xAI at $80 billion and X at $33 billion (a $45 billion valuation less $12 billion in debt). Elon Musk states the companies' futures are "intertwined" and that combining xAI's data, models, compute, and talent with X's distribution and more than 600 million active users will "unlock immense potential," positioning the combined company to build a platform that "doesn't just reflect the world but actively accelerates human progress." The deal follows two years of rapid growth at xAI since its founding and a period in which Musk describes X as having become one of the most efficient companies in the world.
2025 (July 7)	Controversy	Grok is found producing antisemitic responses on X, linking Jewish surnames to negative stereotypes and conspiracy tropes about media and government control. The posts follows Musk's June announcement that Grok would be "retrained" to reduce political correctness. xAI acknowledges the issue, claiming to be removing offensive content and retraining the model. Some posts are deleted, but many remain. The Anti-Defamation League condemns Grok's behavior as "irresponsible and dangerous," warning it amplifies extremist rhetoric and antisemitic ideology already spreading on X.
2025 (July 9)	Release	xAI releases Grok 4, trained using large-scale reinforcement learning on the Colossus supercomputer to refine reasoning at pretraining scale, with a 6x improvement in training compute efficiency over prior methods. The model is trained to natively use tools such as code execution and web/X search, choosing its own search queries to answer complex or real-time questions. A "Grok 4 Heavy" variant introduces parallel test-time compute, evaluating multiple hypotheses simultaneously, and becomes the first model to score 50% on the Humanity's Last Exam benchmark. Grok 4 sets a new state-of-the-art among closed models on ARC-AGI-2 (15.9%) and leads on benchmarks including USAMO 2025, HMMT 2025, and AIME 2025. The Grok 4 API offers a 256,000-token context window with real-time search integration and enterprise security certifications including SOC 2 Type 2, GDPR, and CCPA. An upgraded Voice Mode adds live camera-based visual understanding during voice conversations. Grok 4 launches for SuperGrok and Premium+ subscribers, with a new SuperGrok Heavy tier offering access to Grok 4 Heavy.
2025 (July 9)	Comparison	xAI's own launch materials for Grok 4 detail how it claimed to compare against contemporaneous rivals Claude Opus 4 and Google's Gemini 2.5 Pro. On GPQA, Grok 4 reports 87.5% against Claude Opus 4's 79.6% and Gemini 2.5 Pro's 86.4%. On ARC-AGI-2, Grok 4's 15.9% is presented as nearly double Claude Opus 4's roughly 8.6%, with Gemini 2.5 Pro trailing further at 4.9%. On AIME'25, Grok 4 (with Python) reaches 98.8% against Claude Opus 4's 75.5% and Gemini 2.5 Pro's 88%. On HMMT 2025, Grok 4 (with Python) claims 93.9% against Claude Opus 4's 58.3% and Gemini 2.5 Pro's 82.5%. On the agentic Vending-Bench simulation, xAI reports Grok 4 achieving a net worth of $4,694.15 across five runs, more than double Claude Opus 4's $2,077.41. On Humanity's Last Exam, Grok 4 Heavy becomes the first model to score above 50% on the text-only subset, ahead of Gemini Deep Research, OpenAI's o3, and Gemini 2.5 Pro on the full multimodal set. As with Grok 3's launch comparisons, these figures come directly from xAI's own charts rather than independent third-party evaluation.
2025 (July 14)	Commercialization	xAI announces xAI for Government, a suite of frontier AI products tailored for U.S. government agencies. The initiative brings advanced tools like Grok 4, Deep Search, and Tool Use to federal, state, and national security sectors, aiming to enhance efficiency and innovation in public services, science, and defense. It includes custom AI models for classified environments, engineering support, and specialized applications. xAI also secures a $200 million ceiling contract with the United States Department of Defense, and its products are now available via the General Services Administration schedule.
2025 (July 9 – August)	Personnel	A wave of senior departures hits xAI in the wake of Grok's antisemitic-posting controversy. Linda Yaccarino, chief executive of X, resigns on July 9, one day after Grok posted antisemitic content praising Adolf Hitler, though she gives no specific reason; an analyst suggests the timing reflects "a lack of fit between her approach and Elon Musk's style." Yaccarino had led X since 2023 and previously served as chair of global advertising and partnerships at NBCUniversal; xAI had acquired X in an all-stock deal valuing X at $33 billion in March 2025. Also departing are chief financial officer Mike Liberatore, who had held the role since April and left around the end of July, and senior commercial lawyer Raghu Rao. Liberatore, a former Airbnb executive, had been involved in xAI's $5 billion debt raise and a separate $5 billion strategic equity investment in June. General counsel Robert Keele resigns August 8 after just over a year in the role, citing philosophical differences with Musk and a desire to spend more time with his children, and is replaced by Lily Lim, a former NASA spacecraft-navigation engineer who had joined xAI in late 2024 as a privacy and IP specialist. The departures follow revelations that Grok's antisemitic outputs stemmed from engineers tuning the model's parameters to make it less "politically correct."
2025 (August 5)	Product launch	xAI launches Grok Imagine, an image and video generation tool available to SuperGrok and Premium+ users via the iOS app. The service creates 15-second video clips with sound from text or image prompts, offering a "spicy mode" that allows limited sexualized content. Grok Imagine automatically produces image variations and can animate artwork or photos, including depictions of public figures. Demonstrations include an animated Nikola Tesla image. The launch follows xAI's July release of 3D-animated AI companions for Grok.
2025 (August 7)	Competition	OpenAI launches GPT-5, its "smartest, fastest, most useful model yet," unifying reasoning and non-reasoning capabilities into a single system with a real-time router that decides how much to think based on query complexity. In head-to-head benchmarking published by OpenAI, GPT-5 outperforms Grok 4 across every listed category — AIME 2025 math (94.6% vs. 91.7%), GPQA science reasoning (85.7% vs. 87.5%, Grok 4 slightly ahead on this one metric), LiveCodeBench coding (95.3% vs. 79.0%), SWE-bench Verified (74.9% vs. 55.4% for real-world coding), MMMU multimodal understanding (84.2% vs. 78.3%), and the FrontierMath tiers 1–3 benchmark (26.3% vs. 12.0%), positioning GPT-5 as the new frontier leader across most reasoning, coding, and multimodal tasks at the time of its release.
2025 (August 10)	Competition	ChatGPT defeats Grok in the final of the Kaggle Game Arena chess tournament, a bracket-style competition pitting leading AI models against each other, with Gemini finishing third. OpenAI CEO Sam Altman celebrates the win on X, prompting Elon Musk to respond that Grok 5 will be "much better at chess" and predicting it would "beat any human or AI" by the end of the year. The tournament, run independently by Kaggle, offers a rare direct head-to-head format for comparing frontier models' strategic reasoning outside standard benchmark suites.
2025 (August 18)	Deployment	xAI makes Grok Imagine, its AI-powered image and video generation tool, free to use. Initially available only to SuperGrok and Premium+ subscribers, the feature now allows all users to create short video clips and photorealistic images from text or image prompts directly within the X app. Grok Imagine supports sound, animation, and varied visual styles. The update follows xAI's broader push to make creative AI tools accessible to a wider audience, positioning Grok as a direct competitor to OpenAI's Sora and similar generative video technologies.
2025 (August 21)	Privacy	Hundreds of thousands of user conversations with Grok are unintentionally exposed in Google search results. When users share chat transcripts via Grok, unique links make them publicly searchable, with nearly 300,000 conversations indexed. Transcripts include sensitive content, from medical advice to instructions for illegal activities. Experts warn this represents a "privacy disaster in progress," highlighting risks of personal information exposure. Similar issues had occurred with other AI chatbots. Users are unaware their shared chats would appear publicly, raising concerns about consent and data handling.
2025 (August 21)	Announcement	Elon Musk announces via X (formerly Twitter) that training for Grok 5—the company's next-generation large language model—would commence the following month, in September. This revelation came in a reply thread amid discussions on xAI's broader roadmap, including SpaceX's Starship 10 launch on October 19 and Tesla's Tesla Autopilot V14 rollout. Musk frames the update within a "compelling" long-term strategy, emphasizing Grok 5's potential to advance toward artificial general intelligence (AGI) by surpassing human-level capabilities in AI engineering and reasoning tasks.
2025 (August 28)	Release	xAI launches grok-code-fast-1, a coding-focused reasoning model built from a new architecture and trained on a curriculum of programming-related pretraining data and real-world pull-request datasets, optimized for agentic coding workflows using tools like grep, terminal access, and file editing. The model is offered free for a limited time through launch partners including GitHub Copilot, Cursor (code editor), Cline (software), Roo Code, Kilo Code, opencode, and Windsurf (code editor), and scores 70.8% on SWE-Bench-Verified using xAI's internal harness. It is priced at $0.20 per million input tokens and $1.50 per million output tokens via the xAI API. The model had been released a week earlier in stealth under the codename "sonic."
2025 (August)	Personnel	Igor Babuschkin, xAI's co-founder and chief engineer, announces his departure from the company to launch an investment firm focused on AI safety research. His exit follows a broader wave of senior departures earlier in the summer, including xAI's chief financial officer and general counsel, amid ongoing engineering turnover at the company. According to the Financial Times, departures across Musk's companies, including xAI, X, and Tesla, have been driven by an exhausting work pace, political disagreements with Musk, and staff burnout.
2025 (September 8)	Announcement	Elon Musk states that xAI will match OpenAI, Google, and Anthropic in AI capability by 2026, despite having far fewer employees and having entered the field roughly two years after its rivals. He attributes xAI's rapid progress to a leaner organizational structure and less bureaucracy, framing the company's growth as demonstrating an ability to close the gap with better-resourced, more established competitors.
2025 (September 25)	Commercialization	xAI expands xAI for Government through the General Services Administration's OneGov Initiative, making the platform available to every federal department, agency, or bureau for a nominal $0.42 over an 18-month period. The expansion brings Grok 4 and Grok 4 Fast to federal employees and includes a dedicated team of "Grok Engineers" to support agencies implementing the tools. xAI positions itself as the only AI lab committed to maintaining feature parity between its commercial and government-available models, in alignment with the Donald Trump administration's AI Action Plan.
2025 (September 29)	Deployment	Microsoft and xAI launch Grok 4 on Azure AI Foundry, offering enterprise users a frontier-level reasoning model with safety, flexibility, and control. Built on xAI's Colossus supercomputer, Grok 4 delivers a 128K-token context window, native tool use, and integrated web search. Emphasizing first-principles reasoning and reinforcement learning, it excels at complex math, science, and logic tasks. Azure AI Content Safety is enabled by default, ensuring responsible deployment. The platform also provides specialized variants for reasoning, non-reasoning, and coding. Grok 4 integrates real-time data for dynamic insights and enterprise-ready AI applications.
2025 (October 16)	Competition	Grok Code Fast 1 overtakes ChatGPT, Gemini, and Claude to lead OpenRouter's programming benchmarks and total token usage, signaling a shift in AI adoption among developers. Processing 56.7 billion tokens in Python tasks and 4.83 trillion overall, Grok demonstrates coding efficiency and broad integration. Its success stems from being software-engineering focused, offering free and paid tiers, and emphasizing speed, cost, and usability over sheer size. This surge challenges established models and highlights a trend favoring practical, task-specific AI solutions in real-world development workflows.
2025 (October 17)	Security	Cybercriminals manipulate Grok to spread phishing links and malware on X, affecting millions of users. Researchers at ESET uncover the "Grokking" campaign, which exploits a vulnerability through "prompt injection," embedding hidden instructions in video metadata that Grok interprets and reposts. This causes Grok to unknowingly share fraudulent URLs that steal banking data or install malware. Experts warn the case highlights AI's vulnerability to manipulation and the risks of integrating generative AI with social platforms, urging stricter moderation and real-time security safeguards.
2025 (November 15)	Announcement	In an interview with investor Ron Baron, Elon Musk describes Grok 5 as xAI's largest model to date, a 6-trillion-parameter Mixture-of-Experts model compared to the 3-trillion-parameter architecture underlying Grok 3 and Grok 4, with markedly higher intelligence density per gigabyte of parameters. He states the model is natively multimodal across text, images, video, and audio, with substantially improved tool use, tool creation, and real-time video understanding, which he argues is necessary for artificial general intelligence.
2025 (November 17)	Release	xAI releases Grok 4.1, available to all users across grok.com, X, and the iOS and Android apps, following a two-week silent rollout during which it was preferred over the previous production model 64.78% of the time in blind pairwise evaluations. The model focuses on improving creative, emotional, and collaborative interaction quality using the same reinforcement learning infrastructure as Grok 4, applying frontier agentic reasoning models as reward models to optimize style, personality, and helpfulness. Grok 4.1 Thinking tops the LMArena Text Arena leaderboard, and Grok 4.1 in non-reasoning mode outperforms every other model's full-reasoning configuration on the same leaderboard. It leads on EQ-Bench3 (emotional intelligence) and ranks highly on Creative Writing v3, and its non-reasoning variant roughly halves the hallucination rate of Grok 4 Fast on information-seeking queries.
2025 (November 17)	Comparison	xAI's own launch materials for Grok 4.1 detail how it claimed to compare against contemporaneous rivals from Anthropic, Google, and OpenAI, though notably the leaderboard snapshot predates Anthropic's Claude Opus 4.5 and shows only Claude Opus 4.1 and Claude Sonnet 4.5. On the LMArena Text Leaderboard, Grok 4.1 Thinking holds the top position at 1483 Elo, ahead of Google's Gemini 2.5 Pro (1452), Claude Sonnet 4.5 Thinking (1450), Claude Opus 4.1 Thinking (1449), Claude Sonnet 4.5 (1445), Claude Opus 4.1 (1440), and OpenAI's GPT-5 High (1438) and o3 (1437); Grok 4.1's non-reasoning variant ranks second at 1465, ahead of every rival's full-reasoning configuration. On EQ-Bench3, an emotional-intelligence benchmark, Grok 4.1 Thinking scores 1586 against Google's Gemini 2.5 Pro (1460), OpenAI's GPT-5 Chat (1364), and Anthropic's Claude Opus 4 (1304). On Creative Writing v3, Grok 4.1 Thinking scores 1721.9, behind an early preview build of OpenAI's GPT-5.1 (reported as "Polaris Alpha," 1756.2) but ahead of OpenAI's o3 (1696.4) and Claude Sonnet 4.5 (1648.7). As with Grok 3 and Grok 4's launch comparisons, these figures come directly from xAI's own charts rather than independent third-party evaluation, though the LMArena figures reflect a public, third-party-hosted leaderboard rather than an internal xAI benchmark.
2025 (November 19)	Commercialization	xAI signs a framework agreement with the Kingdom of Saudi Arabia and HUMAIN, an AI company owned by the kingdom's Public Investment Fund, to design and operate hyperscale GPU data centers in the country and deploy Grok nationwide across public and private sectors. Grok is also set to integrate into HUMAIN's agent platform, HUMAIN ONE, providing government, enterprise, and consumer users with AI copilots and autonomous workflows. Elon Musk describes it as the first partnership to deploy Grok across an entire country.
2025 (November 19)	Release	xAI launches Grok 4.1 Fast, a tool-calling model with a 2-million-token context window, alongside the Agent Tools API, giving agents access to real-time X data, web search, remote code execution, file retrieval, and external MCP servers on xAI's own infrastructure. Trained via reinforcement learning across simulated multi-domain environments and long-horizon, multi-turn scenarios, the model tops the τ²-bench Telecom customer-support benchmark and the Berkeley Function Calling v4 benchmark for tool-calling accuracy, and outperforms GPT-5, Claude Sonnet 4.5, and Gemini 3 Pro on several agentic search benchmarks. It cuts the hallucination rate roughly in half compared to Grok 4 Fast while matching Grok 4's factuality on FActScore. The model releases in reasoning and non-reasoning variants, priced at $0.20 per million input tokens and $0.50 per million output tokens.
2025 (December 11)	Deployment	xAI announces a partnership with the Government of El Salvador to launch what it describes as the world's first nationwide AI-powered education program, deploying Grok across more than 5,000 public schools over two years to provide adaptive, curriculum-aligned tutoring to over one million students. President Nayib Bukele frames the initiative as part of a broader strategy of El Salvador adopting bold technology policy ahead of other nations. xAI states the partnership will generate methodologies, datasets, and frameworks intended to inform responsible AI use in classrooms more broadly.
2025 (December 17)	Release	xAI launches the Grok Voice Agent API, opening the voice technology already powering Grok Voice in its mobile apps and Tesla vehicles to all developers. Built entirely in-house, including custom voice activity detection, tokenizer, and audio models, it ranks first on the Big Bench Audio reasoning benchmark with an average time-to-first-audio under one second, about five times faster than the nearest competitor. The API supports dozens of languages with automatic mid-conversation language switching and rates ahead of OpenAI's Realtime API in blind human evaluations across most tested languages. Developers can integrate custom tools or xAI's real-time web and X search, and access multiple expressive voices supporting auditory cues like `[whisper]` and `[laugh]`. Pricing is a flat $0.05 per minute of audio, undercutting Deepgram, ElevenLabs, OpenAI's Realtime API, and Bland AI. The API is compatible with the OpenAI Realtime API specification and available via an official LiveKit plugin.
2025 (December 22)	Commercialization	xAI is selected by the United States Department of War to supply Frontier AI systems as part of its GenAI.mil initiative, giving up to 3 million military and civilian department employees access to xAI technology at Department of War Impact Level 5 (IL5). The partnership, coordinated with the department's Chief Digital and Artificial Intelligence Office, brings xAI for Government's models, agentic tools, research platform, and API to both enterprise administrative use cases and, through a separate long-term partnership, a family of government-optimized foundation models for classified operational workloads. Department of War users also gain access to real-time insights from the X platform.
2025 (December 22)	Release	xAI launches the Collections API, a retrieval-augmented generation (RAG) system built into the xAI API, letting developers upload and search datasets such as PDFs, spreadsheets, and codebases without managing separate indexing infrastructure. The system uses OCR and layout-aware parsing to preserve document structure, supports semantic, keyword, and hybrid search with reranking, and automatically reindexes files when updated. On internal benchmarks, xAI's Collections API paired with Grok 4.1 Fast outperforms Google's File Search with Gemini Pro 3 and OpenAI's Vector Store with GPT-5.1 on finance and coding retrieval tasks, and performs comparably on legal reasoning tasks (LegalBench). User data stored in Collections is not used for model training without explicit consent. Retrieval is priced at $2.50 per 1,000 searches, with indexing and storage free for an initial trial period.
2025 (December 30)	Product launch	xAI launches Grok Business and Grok Enterprise, bringing team-oriented access to Grok with built-in security and privacy guarantees, including no training on customer data. Grok Business, priced at $30 per seat per month, offers self-serve onboarding for small-to-medium teams with team management, consolidated billing, and usage analytics. Grok Enterprise adds custom single sign-on, directory sync (SCIM), and advanced audit and security controls for larger organizations. Grok gains permission-aware access to company knowledge starting with Google Drive via "Apps," providing citations linking directly to source documents, and can perform agentic search across large document stores through a Collections API in Projects. An optional Enterprise Vault tier offers a dedicated, isolated data plane with application-level encryption and customer-managed encryption keys for the most security-sensitive customers.
2026 (January 6)	Commercialization	xAI completes an upsized Series E funding round, raising $20 billion against an initial $15 billion target, with participation from Valor Equity Partners, StepStone Group, Fidelity Management & Research Company, the Qatar Investment Authority, MGX, and Baron Capital Group, alongside strategic investors Nvidia and Cisco Systems Investments. xAI reports ending 2025 with more than one million H100 GPU equivalents across its Colossus I and II supercomputers and approximately 600 million monthly active users across X and the Grok apps. The company states Grok 5 is in training and the funding will accelerate infrastructure buildout and product development.
2026 (January 10)	Regulation	Indonesia becomes the first country to block Grok, citing risks from non-consensual, AI-generated pornographic deepfakes. The government states such content violates human rights, dignity, and digital security, particularly for women and children. The ban follows mounting international criticism and Grok's partial restriction to paying users on X. Officials argue these measures are insufficient, while Musk downplays concerns and faces potential fines amid growing global regulatory pushback.
2026 (January 12)	Regulation	Malaysia blocks access to Grok after it fails to comply with regulatory orders addressing the creation of nonconsensual sexualised images. The Malaysian Communications and Multimedia Commission states that xAI and X did not adequately mitigate inherent design risks, relying instead on user reporting. The ban follows Indonesia's earlier prohibition and comes amid international condemnation over sexually explicit deepfakes, including depictions of minors, generated by Grok's image tools.
2026 (January 15)	Restriction	X announces that Grok would stop editing images of real people to produce sexualised or revealing depictions in jurisdictions where such content is illegal, following public backlash and regulatory pressure. UK officials and Ofcom welcome the move but continue investigations into possible legal breaches. Campaigners and victims argue the safeguards arrive too late to prevent harm, while experts raise ongoing concerns about enforcement, geoblocking effectiveness, and accountability for AI-generated abuse.
2026 (January 22)	Controversy	Researchers estimate that Grok's image-generation tool produced about three million sexualised images in 11 days, including roughly 23,000 appearing to depict children. The findings, published by the Center for Countering Digital Hate, describe Grok as enabling industrial-scale creation of non-consensual sexual imagery. The feature, launched in late December 2025, had gone viral on X before restrictions were imposed on paid users. The report intensifies international criticism and prompts regulatory action in multiple countries.
2026 (January)	Litigation	Four plaintiffs — three women and one man, identified pseudonymously as "South Carolina Roe," "South Carolina Doe," "New Jersey Doe," and "Ohio Doe" — file a class-action lawsuit against xAI in the U.S. District Court for the Northern District of California in San Jose, seeking to represent hundreds of people whose photos Grok allegedly altered into non-consensual sexualized images and explicit videos. The suit alleges Grok "humiliates and sexually exploits women, girls, and others," and that it produced illegal child sexual abuse material in at least one case. The suit follows the December 2025 rollout of Grok's "edit image" feature, which prompted widespread reports of misuse to create such content, alongside international backlash, other lawsuits, and a since-opened investigation by the State of California. Elon Musk is not named as a defendant, but the suit alleges he and xAI "have chosen to capitalize on predators' insatiable appetite for non-consensual sexual images and videos of real people, including children."
2026 (January 28)	Integration	Grok Imagine API is introduced as xAI's unified video-audio generation and editing platform, designed to support end-to-end creative workflows. It enables text-to-video and image-to-video generation, as well as advanced video editing functions such as restyling scenes, controlling motion, and adding or removing objects. xAI positions the model as competitive with leading providers on quality while emphasizing lower latency and cost. Benchmark comparisons report strong instruction following and consistency, targeting developers, creative teams, and enterprise use cases.
2026 (January)	Personnel	Greg Yang, one of xAI's founding members, announces he is stepping back from his role to focus on his health.
2026 (February 2)	Acquisition	SpaceX acquires xAI, merging the AI company with SpaceX's rockets, Starlink satellite internet, and X platform into a single vertically-integrated organization, in a transaction CNBC reports as the largest merger in history, valuing SpaceX at $1 trillion and xAI at $250 billion. Elon Musk frames the acquisition around a long-term vision of space-based AI compute, arguing that terrestrial power and cooling constraints make large-scale AI training and inference on Earth unsustainable, and proposes launching a constellation of orbital data centers powered by direct solar energy, enabled by Starship's payload capacity. Following the acquisition, xAI's products and communications increasingly appear under the "SpaceXAI" name.
2026 (February 3)	Regulation	Regulators in the United Kingdom and European Union intensify scrutiny of Grok over its alleged role in generating sexualized deepfake images without consent. The Information Commissioner's Office opens an investigation into Grok's handling of personal data, citing potential privacy and safety violations, while coordinating with Ofcom. Separately, the European Commission announces a probe into xAI, Grok's parent company, amid broader concerns about safeguards against harmful AI-generated content.
2026 (February 3)	Commentary	An article presents Grok as the most provocative and unstable of major AI assistants, deliberately shaped by Elon Musk to reject "political correctness" and adopt an edgy, confrontational tone. Marketed as "maximally truth-seeking," Grok instead had sparked repeated controversies, including antisemitic outputs, conspiracy claims, and the mass generation of non-consensual sexualised images. Unlike Claude's ethics-driven constitution or ChatGPT's people-pleasing optimism, Grok's loose behavioral constraints allow sharper satire, insults, and boundary-pushing responses. Critics argue this design amplifies harm and misinformation, while supporters see it as a rebellious alternative. The article frames Grok as a cautionary example of how tuning AI "character" can produce real-world consequences.
2026 (February 10)	Personnel	Tony Wu, another of xAI's founding members, announces his resignation from the company in a post on X, writing "It's time for my next chapter" and thanking Elon Musk "for the ride of a lifetime," becoming the latest in a series of co-founder departures that also includes Igor Babuschkin, Kyle Kosic, and Christian Szegedy. Musk had launched xAI in 2023 with 11 other co-founders. Wu's departure comes as xAI faces consumer backlash and regulatory investigations in multiple countries over Grok's role in generating non-consensual sexualised deepfakes, including of children.
2026 (February 17)	Deployment	Tesla announces the rollout of Grok to vehicle infotainment systems in the U.K. and several European markets, expanding its in-car AI features amid ongoing regulatory scrutiny. The move follows a 27% decline in Tesla's European sales, despite strong regional growth in electric vehicles. Regulators investigate Grok over safety, content moderation, and compliance with the Digital Services Act. Experts also raise concerns about driver distraction and minors' access to the chatbot.
2026 (February 17)	Announcement	Elon Musk states that Grok 4.2 is expected to be roughly an order of magnitude smarter and faster than Grok 4 upon completion of its public beta in the following month. He adds that ongoing bug fixes and daily improvements are informed by critical feedback from beta users.
2026 (February 24)	Adoption	The United States Department of Defense signs an agreement with xAI to deploy Grok in classified military systems, marking the first alternative to Anthropic's Claude at the highest security level. The move follows a dispute with Anthropic over safeguards limiting mass surveillance and autonomous weapons. xAI accepts the Pentagon's "all lawful purposes" requirement, while talks continue with Google and OpenAI over classified access, highlighting growing tensions over AI governance, ethics, and military integration.
2026 (March 13)	Controversy	Elon Musk publicly acknowledges that xAI's AI coding tools, including Grok-based products, are not competitive with Anthropic's Claude Code or OpenAI's Codex (software), stating the underlying system needs to be rebuilt. The admission follows a $2 billion investment by Tesla in xAI's Series E round at an approximate $230 billion valuation, and adds weight to a shareholder lawsuit accusing Musk of directing Tesla capital into his own private venture.
2026 (March 24)	Litigation	The mayor and city council of Baltimore file a lawsuit against xAI in the circuit court for Baltimore City, alleging Grok violated consumer protection law by generating non-consensual sexualised images and child sexual abuse material without disclosing the risks to users. The suit argues xAI deceptively marketed Grok as a general-purpose AI assistant and X as a mainstream social platform, and that any photo uploaded by a Baltimore resident, including of their children, risked being transformed into sexualised deepfakes without consent. Unlike prior suits brought by individual plaintiffs, Baltimore's case is based on municipal ordinance and consumer protection violations. Mayor Brandon Scott says the city will not allow the platform to enable child sexual exploitation. Elon Musk had stated in January 2026 that he was unaware of any underage images generated by Grok, before xAI added image-generation restrictions following international backlash.
2026 (March 26)	Litigation	The District Court of Amsterdam (Rechtbank Amsterdam) issues an immediate ban prohibiting X and X.AI from generating and distributing deepfake sexual images of people, including minors, residing in the Netherlands, ruling in summary proceedings brought by online-abuse expertise centre Offlimits with support from the Netherlands Victim Support Fund. The case follows X's late-2025 introduction of an image-editing feature within Grok that led to mass generation of non-consensual sexualised deepfakes, including child sexual abuse material, and had already prompted a formal European Commission investigation under the Digital Services Act. The court finds X.AI qualifies as a data controller under the GDPR for Dutch residents' personal data, giving it jurisdiction, and rules that X and X.AI failed to demonstrate their safeguards — including input filters intended to reject CSAM-related requests — were actually effective, with Offlimits presenting screenshots showing sexualised deepfakes remained generable even after additional safeguards were introduced. The ruling bans generating or distributing non-consensual undressing imagery of Dutch residents and any CSAM within the Netherlands, and prohibits X from offering Grok's image-editing functionality on its platform at all so long as it remains exploitable for illicit content. X and X.AI must confirm compliance in writing, with a penalty of €100,000 per day, capped at €10 million, for non-compliance.
2026 (March 28)	Personnel	Manuel Kroiss and Ross Nordeen, the last of xAI's original eleven co-founders, depart the company, completing a total exodus of the founding team. The cohort — which included Jimmy Ba (co-author of the widely cited Adam (optimization algorithm) optimization paper), Igor Babuschkin, Christian Szegedy, Tony Wu, Greg Yang, Toby Pohlen, Zihang Dai, Guodong Zhang, and Kyle Kosic — had been recruited by Elon Musk in 2023 from labs including Google DeepMind, Google, Microsoft, and OpenAI. The departures accelerate sharply after Tony Wu's February 10, 2026 resignation, with Jimmy Ba departing within 24 hours reportedly amid tensions over demands to improve model performance.
2026 (March 31)	Commercialization	Following the February 2026 SpaceX–xAI merger, which valued SpaceX at approximately $1 trillion and xAI at $250 billion for a combined $1.25 trillion, financial reporting indicates SpaceX is preparing what would be the largest initial public offering in history, targeting a market capitalization of $1.75 trillion or more and aiming to raise between $40 billion and $80 billion — up sharply from SpaceX's standalone $400 billion valuation in July 2025. The company reportedly intends to allocate a substantial portion of IPO proceeds to retail investors. SpaceX reported $16 billion in revenue for 2025; combined with xAI's revenue, analysts estimate the merged entity's total 2025 revenue at roughly $20 billion, a figure that would imply a steep valuation multiple relative to comparable IPOs if the $1.75 trillion target is reached.
2026 (March)	Litigation	Tesla shareholders sue Elon Musk for breach of fiduciary duty over Tesla's $2 billion investment in xAI's Series E funding round, alleging Musk directed Tesla shareholder capital into his own private AI venture. The suit gains additional weight after Musk's March 13, 2026 acknowledgment that xAI's coding tools were uncompetitive.
2026 (March)	Litigation	Three Tennessee teenage plaintiffs, referred to as Jane Does, file a class-action lawsuit against xAI alleging its Grok chatbot was used to generate child sexual abuse material (CSAM) from photographs taken while they were underage, accusing the company of producing CSAM, benefiting from sex trafficking ventures, negligence, defective product design, and creating a public nuisance. Attorney Annika Martin, representing the plaintiffs, states the suit's public-nuisance framing reflects that the harm "is not just something that is problematic for our clients... this is something that is a scourge on society," and that the plaintiffs seek both stronger guardrails and monetary compensation. The suit alleges perpetrators — including a plaintiff's male friend — used Grok to alter existing photographs into sexually explicit imagery. The case would later expand in July 2026 to add plaintiffs from Wyoming and Wisconsin and to name Stability AI as a co-defendant, following the SpaceX–xAI merger and the company's rebranding as SpaceXAI.
2026 (April 9)	Personnel	Anthony Armstrong, a former Morgan Stanley banker who advised Elon Musk during the 2022 acquisition of X and was named CFO of the combined xAI and X in October 2025, departs the company as part of a continuing wave of senior exits, according to The Information. Armstrong had reported to Bret Johnsen, finance chief of the merged xAI–SpaceX entity, and had been responsible for steering X's finances after an advertiser exodus following relaxed content moderation.
2026 (April 17)	Release	xAI launches standalone Speech to Text (STT) and Text to Speech (TTS) APIs, built on the same underlying stack powering Grok Voice, Tesla vehicles, and Starlink customer support. STT offers batch and real-time transcription with word-level timestamps, speaker diarization, multichannel support, and inverse text normalization for numbers, dates, and currencies, priced at $0.10 per hour for batch and $0.20 per hour for streaming, undercutting competitors including AssemblyAI, ElevenLabs, and Deepgram. TTS supports fine-grained delivery control via speech tags such as `[laugh]`, `[whisper]`, and `<emphasis>`, priced at $15.00 per million characters, also below competing providers including OpenAI, InWorld AI, Cartesia (company), and ElevenLabs. Both APIs support more than 25 languages.
2026 (April 23)	Release	xAI releases `grok-voice-think-fast-1.0`, its flagship voice agent model, via API, built through close collaboration with Starlink for complex, ambiguous, multi-step voice workflows in customer support and sales. The model reasons in the background without adding response latency, precisely collects and confirms structured data such as addresses and account numbers even with accents or disfluencies, and is designed to reason through edge cases rather than default to confident but incorrect answers. It tops the τ-voice Bench leaderboard for full-duplex voice agents under realistic conditions, natively supporting more than 25 languages. The model powers Starlink's phone sales and support line, achieving a 20% sales conversion rate and a 70% autonomous resolution rate for support inquiries across 28 distinct tools.
2026 (April 30)	Product launch	xAI introduces Custom Voices, letting users clone their own voice from about a minute of recorded audio for use across the Grok Text to Speech and Voice Agent APIs, with no additional charge. Cloning takes under two minutes and the resulting voice inherits full text-to-speech capabilities, including speech tags, multilingual output, and streaming. To prevent unauthorized cloning, xAI requires a two-stage verification process: the speaker reads a passphrase transcribed and matched in real time by xAI's speech-to-text engine, and speaker embeddings from that passphrase and the full recording are compared to confirm they come from the same person, preventing voices from being cloned from pre-existing recordings or without the speaker's live participation.
2026 (May 6)	Commercialization	xAI signs an agreement with Anthropic to provide access to Colossus 1, an AI supercomputer built by xAI featuring over 220,000 Nvidia GPUs, including dense deployments of H100, H200, and next-generation GB200 accelerators. Anthropic plans to use the additional compute to expand capacity for Claude Pro and Claude Max subscribers. As part of the agreement, Anthropic also expresses interest in partnering with xAI and SpaceX to develop orbital AI compute capacity, arguing that space-based compute could offer near-limitless sustainable power with less environmental impact than terrestrial data centers, an area where xAI positions SpaceX's launch cadence and orbital operations experience as uniquely suited.
2026 (May 5)	Release	xAI releases Grok 4.3 via the xAI API, describing it as the company's fastest and most intelligent model to date. The model tops Artificial Analysis leaderboards for agentic tool calling and instruction following, and ranks first on Vals AI's enterprise-domain benchmarks including case law and corporate finance. Grok 4.3 supports a 1-million-token context window and is priced at $1.25 per million input tokens and $2.50 per million output tokens.
2026 (May 6)	Product launch	xAI launches Connectors, deep integrations bringing external apps and services directly into Grok on web, iOS, and Android. Available connectors at launch include SharePoint, Microsoft Outlook, OneDrive, Google Workspace, Notion (productivity software), GitHub, and Linear (software), letting Grok search, read, and — with permissions enabled — write and edit documents, emails, calendar events, code, and project tasks across these services. xAI also launches Bring Your Own MCP, allowing users to connect any custom Model Context Protocol server, including homegrown knowledge bases, proprietary APIs, or internal MCP gateways.
2026 (May 6)	Release	xAI launches Quality Mode for the Grok Imagine API, an enterprise-focused image generation and editing model emphasizing photorealism, multilingual text rendering, and tighter prompt-following for consistent brand results. On the independent Text-to-Image Arena leaderboard, xAI's Grok Imagine API ranks third among labs behind OpenAI and Google, ahead of Luma AI and Microsoft AI, and above xAI's own previous model. xAI positions the model for enterprise use cases such as product visualization, branded marketing assets, and UGC-style content generation, and pairs it with Grok's video capabilities for social and advertising video assets.
2026 (May 15)	Integration	xAI makes Grok available inside Hermes Agent, Nous Research's open-source, self-improving personal agent that runs persistently on user-owned hardware and builds long-term memory across sessions. Subscribers on any tier can connect their Grok account to use Grok 4.3 for reasoning and conversation, Grok Text-to-Speech for spoken responses, and Grok Imagine for image and video generation within Hermes, which also connects to messaging platforms including WhatsApp, Discord, Telegram (software), and Signal (software).
2026 (April 9)	Personnel	Anthony Armstrong, a former Morgan Stanley banker who advised Elon Musk during the 2022 acquisition of X and was named CFO of the combined xAI and X in October 2025, departs the company as part of a continuing wave of senior exits, according to The Information. Armstrong had reported to Bret Johnsen, finance chief of the merged xAI–SpaceX entity, and had been responsible for steering X's finances after an advertiser exodus following relaxed content moderation.
2026 (April 17)	Release	xAI launches standalone Speech to Text (STT) and Text to Speech (TTS) APIs, built on the same underlying stack powering Grok Voice, Tesla vehicles, and Starlink customer support. STT offers batch and real-time transcription with word-level timestamps, speaker diarization, multichannel support, and inverse text normalization for numbers, dates, and currencies, priced at $0.10 per hour for batch and $0.20 per hour for streaming, undercutting competitors including AssemblyAI, ElevenLabs, and Deepgram. TTS supports fine-grained delivery control via speech tags such as `[laugh]`, `[whisper]`, and `<emphasis>`, priced at $15.00 per million characters, also below competing providers including OpenAI, InWorld AI, Cartesia (company), and ElevenLabs. Both APIs support more than 25 languages.
2026 (April 23)	Release	xAI releases `grok-voice-think-fast-1.0`, its flagship voice agent model, via API, built through close collaboration with Starlink for complex, ambiguous, multi-step voice workflows in customer support and sales. The model reasons in the background without adding response latency, precisely collects and confirms structured data such as addresses and account numbers even with accents or disfluencies, and is designed to reason through edge cases rather than default to confident but incorrect answers. It tops the τ-voice Bench leaderboard for full-duplex voice agents under realistic conditions, natively supporting more than 25 languages. The model powers Starlink's phone sales and support line, achieving a 20% sales conversion rate and a 70% autonomous resolution rate for support inquiries across 28 distinct tools.
2026 (April 30)	Product launch	xAI introduces Custom Voices, letting users clone their own voice from about a minute of recorded audio for use across the Grok Text to Speech and Voice Agent APIs, with no additional charge. Cloning takes under two minutes and the resulting voice inherits full text-to-speech capabilities, including speech tags, multilingual output, and streaming. To prevent unauthorized cloning, xAI requires a two-stage verification process: the speaker reads a passphrase transcribed and matched in real time by xAI's speech-to-text engine, and speaker embeddings from that passphrase and the full recording are compared to confirm they come from the same person, preventing voices from being cloned from pre-existing recordings or without the speaker's live participation.
2026 (May 6)	Commercialization	xAI signs an agreement with Anthropic to provide access to Colossus 1, an AI supercomputer built by xAI featuring over 220,000 Nvidia GPUs, including dense deployments of H100, H200, and next-generation GB200 accelerators. Anthropic plans to use the additional compute to expand capacity for Claude Pro and Claude Max subscribers. As part of the agreement, Anthropic also expresses interest in partnering with xAI and SpaceX to develop orbital AI compute capacity, arguing that space-based compute could offer near-limitless sustainable power with less environmental impact than terrestrial data centers, an area where xAI positions SpaceX's launch cadence and orbital operations experience as uniquely suited.
2026 (May 5)	Release	xAI releases Grok 4.3 via the xAI API, describing it as the company's fastest and most intelligent model to date. The model tops Artificial Analysis leaderboards for agentic tool calling and instruction following, and ranks first on Vals AI's enterprise-domain benchmarks including case law and corporate finance. Grok 4.3 supports a 1-million-token context window and is priced at $1.25 per million input tokens and $2.50 per million output tokens.
2026 (May 6)	Product launch	xAI launches Connectors, deep integrations bringing external apps and services directly into Grok on web, iOS, and Android. Available connectors at launch include SharePoint, Microsoft Outlook, OneDrive, Google Workspace, Notion (productivity software), GitHub, and Linear (software), letting Grok search, read, and — with permissions enabled — write and edit documents, emails, calendar events, code, and project tasks across these services. xAI also launches Bring Your Own MCP, allowing users to connect any custom Model Context Protocol server, including homegrown knowledge bases, proprietary APIs, or internal MCP gateways.
2026 (May 6)	Release	xAI launches Quality Mode for the Grok Imagine API, an enterprise-focused image generation and editing model emphasizing photorealism, multilingual text rendering, and tighter prompt-following for consistent brand results. On the independent Text-to-Image Arena leaderboard, xAI's Grok Imagine API ranks third among labs behind OpenAI and Google, ahead of Luma AI and Microsoft AI, and above xAI's own previous model. xAI positions the model for enterprise use cases such as product visualization, branded marketing assets, and UGC-style content generation, and pairs it with Grok's video capabilities for social and advertising video assets.
2026 (May 15)	Integration	xAI makes Grok available inside Hermes Agent, Nous Research's open-source, self-improving personal agent that runs persistently on user-owned hardware and builds long-term memory across sessions. Subscribers on any tier can connect their Grok account to use Grok 4.3 for reasoning and conversation, Grok Text-to-Speech for spoken responses, and Grok Imagine for image and video generation within Hermes, which also connects to messaging platforms including WhatsApp, Discord, Telegram (software), and Signal (software).
2026 (May 18)	Product launch	xAI launches Skills, a system letting Grok retain taught preferences, formatting rules, and workflow steps persistently across conversations. Every account ships with built-in xAI skills for generating and editing Word documents, presentations, spreadsheets, and PDFs, alongside a "Skill Creator" for building custom skills through conversation; user-created skills override the built-in defaults. Grok can generate production-ready office documents from a single prompt, restructure or polish uploaded files, and save a refined format or workflow as a reusable skill. Skills launch on Grok 4.3 across grok.com, iOS, and Android.
2026 (May 19)	Integration	xAI makes Grok available inside OpenClaw, an open-source, local-first personal assistant agent that runs on user-owned hardware (from a Mac Mini to a Raspberry Pi) and maintains persistent memory across sessions. SuperGrok and X Premium subscribers on any tier can connect their account and interact with a Grok-powered OpenClaw agent through messaging platforms including WhatsApp, Telegram, Slack (software), Discord, Signal, and iMessage.
2026 (May 21)	Integration	xAI makes Grok available inside OpenCode, an open-source terminal-based coding agent, letting SuperGrok and X Premium subscribers connect their account via OAuth to code with the same underlying model that powers Grok Build.
2026 (May 25)	Product launch	xAI launches an early beta of Grok Build, a coding agent and command-line interface for professional software engineering, available to SuperGrok and X Premium Plus subscribers. Grok Build supports a plan-review-approve workflow where users can approve, comment on, or rewrite a proposed plan before execution, with changes shown as diffs. It integrates with existing `AGENTS.md` conventions, plugins, hooks, skills, and MCP servers automatically, and delegates larger tasks to specialized subagents that run in parallel, including in separate Git worktrees. A headless mode (`-p`) supports running agents inside scripts and automations, and the CLI provides Agent Client Protocol support for building custom bots and orchestration apps.
2026 (May 27)	Integration	xAI makes Grok available inside Kilo Code, an open-source agentic engineering platform for Visual Studio Code, JetBrains IDEs, and the terminal offering specialized planning, coding, debugging, and orchestration modes across 500+ supported models. X Premium+ and SuperGrok subscribers can connect their account via OAuth to use Grok models, including Grok Build, within Kilo Code without a separate API key, across the IDE extension, CLI, and web surface.
2026 (May 29)	Release	xAI releases `grok-build-0.1`, a coding model trained specifically for agentic coding tasks including web development, debugging, and MCP support, via the xAI API in public beta. It is the same model powering the Grok Build CLI, served at over 100 tokens per second and priced at $1 per million input tokens and $2 per million output tokens. The model is designed to perform best within agentic coding harnesses such as Grok Build, Cursor, and several third-party tools, and is also available through OpenRouter and the Vercel AI Gateway.
2026 (June 1)	Integration	Composer 2.5, described as a fast model suited to long-running tasks and complex instruction-following, becomes available as a selectable model inside Grok Build via the `/model` menu.
2026 (June 3)	Release	xAI previews `grok-imagine-video-1.5-preview`, its latest image-to-video model, via the xAI API. The model animates a single still image into cinematic video based on a natural-language prompt describing camera movement, pacing, and sound design, while preserving the detail and lighting of the source frame rather than reinterpreting it, at resolutions up to 720p. It also supports staging and chaining multiple animated frames into longer sequences with a consistent visual style.
2026 (June 3)	Integration	xAI partners with Vapi to make Grok Voice the default text-to-speech engine for Vapi's 12 core voices, powering more than 2.5 million voice agents built on the platform. Vapi's independent blind evaluation ranks Grok Voice first among competing providers, and a companion poll of over 4,500 X users on distinguishing a Grok voice clone from the human original splits roughly 50/50. Grok Speech-to-Text and Text-to-Speech become available in the Vapi dashboard, with additional customization, including custom voice cloning, available directly through the Grok Voice API.
2026 (June 9)	Integration	Gopuff and xAI launch Go, an AI-powered personal shopping assistant built into the Gopuff app, combining Grok's text, voice, and image models with Gopuff's demand intelligence drawn from over a decade of order history. Go builds personalized shopping carts in advance using past preferences and external signals such as weather, learns from every order to improve future recommendations, and includes a visual shopping feed powered by Grok Imagine that generates realistic scenes from Gopuff inventory to help visualize carts. It also draws on real-time signals from X and the web via xAI's API tools. Go launches on iOS and Android in the United States, with a UK rollout to follow.
2026 (June 10)	Integration	eToro integrates xAI models into Tori, its AI investing agent, embedding real-time market sentiment drawn from X directly into Tori's workflow. The integration lets eToro's more than 40 million registered users across 75 countries ask Tori to read shifting market moods and analyze live signals using xAI's text models, built on the same API available to any developer through the xAI console.
2026 (June 11)	Regulation	Canada's Privacy Commissioner Philippe Dufresne finds that xAI and X Corp. violated federal privacy law by launching Grok's image-generation tool without adequate safeguards, enabling the creation and distribution of non-consensual sexualised deepfakes targeting women and children. The investigation, opened in January 2026, concludes the feature was deployed without proper risk assessment or sufficient technical controls. During the investigation, xAI and X introduce additional safeguards, including detection systems and new usage restrictions, but Dufresne calls for further improvements, ongoing independent audits, and quarterly compliance reporting. He also argues Canada's privacy framework lacks sufficient enforcement powers, such as the ability to issue binding orders or meaningful financial penalties, and recommends mandatory privacy impact assessments for high-risk AI systems.
2026 (June 11)	Integration	xAI launches the Grok Build Plugin Marketplace, a built-in catalog of installable plugins bundling skills, slash commands, agents, hooks, MCP servers, and language server protocols into single packages. The marketplace opens with partner plugins from MongoDB, Vercel, Sentry, Chrome DevTools, and Cloudflare, browsable and installable from within Grok Build via `/marketplace` or the `grok plugin` command line interface. Every remote plugin is pinned to a specific commit hash, verified at install time, and the catalog is open for developers to submit their own plugins.
2026 (June 15)	Integration	xAI makes Grok available inside Warp, an agentic terminal-based development environment used by nearly one million developers, for existing Grok and X Premium subscribers. Access is enabled by connecting a SuperGrok subscription in Warp's Agent settings and selecting the `grok-build-0.1` model, the same model powering the Grok Build CLI.
2026 (June 15)	Integration	xAI adds an Agent Dashboard to Grok Build, letting users manage multiple coding sessions from a single screen, sorted by state so sessions awaiting input surface first. Users can peek at a session's latest output and reply without leaving the dashboard, dispatch new sessions with configurable models and approval settings, group sessions by working directory, and take over any session's full conversation before returning to the dashboard. Sessions continue running when the dashboard is closed. The feature ships with Grok Build, accessible via `grok dashboard` from the shell or `/dashboard` from within any session.
2026 (June 16)	Release	xAI releases Grok Imagine Video 1.5, its latest image-to-video model, generally available via the Imagine API, with a faster "Video 1.5 Fast" variant rolled out on grok.com/imagine and the iOS and Android apps. The update improves audio and speech generation (sound effects, ambience, and dialogue generated in the same pass, better lip sync), motion and physics consistency over the length of a clip, and generation speed, with Video 1.5 Fast producing 6-second 720p clips in about 25 seconds, down from over 40 seconds previously. xAI also rolls out workflow features including Projects for organizing work, the ability to run multiple generation agents in parallel, and search across a user's generated image and video library.
2026 (June 16)	Integration	xAI launches a free Grok add-in for Microsoft PowerPoint, letting users generate full slide decks from an outline, complete with web or X research, diagrams, and images, or add individual slides, apply styles and themes, and restructure sections with a single instruction. Grok asks clarifying questions as needed while building a deck, and with connectors enabled, can draft slides from recent emails or files in SharePoint or Google Drive. It completes xAI's Microsoft 365 add-in suite alongside the existing Word and Excel add-ins.
2026 (June 17)	Deployment	xAI makes Grok 4.3 generally available on Amazon Bedrock, AWS's managed AI model platform. xAI reports Grok 4.3 achieves the lowest hallucination rate among frontier models, ranking #1 on the Artificial Analysis Omniscience and Tau2 Telecom benchmarks and on Vals AI's Case Law and Corporate Finance benchmarks. The model offers a 1-million-token context window, configurable reasoning effort levels (none, low, medium, high), and is priced at $1.25 per million input tokens and $2.50 per million output tokens, available to developers across supported AWS regions.
2026 (June 18)	Deployment	xAI makes Grok models natively available on Databricks Agent Bricks, Databricks' developer agent platform, as part of the Databricks 2026 Data + AI Summit. The partnership lets engineering teams build AI agents that operate on large volumes of data governed within Databricks' Lakehouse platform, alongside other frontier and open-source models. The move follows Grok's earlier availability on Amazon Bedrock, giving enterprises additional options to run Grok models where their data is already stored.
2026 (June 18)	Integration	xAI launches a free Grok add-in for Microsoft Word, letting users turn prompts and rough notes into structured, long-form documents, draft proposals or reference guides, and rewrite text for clarity, concision, or consistent style across multiple authors. The add-in can search the web or X to bring research and data directly into a document, generate diagrams, and, with connectors enabled, draft from recent emails or files in SharePoint or Google Drive. It complements xAI's existing Excel and PowerPoint add-ins.
2026 (June 22)	Integration	xAI introduces `/goal` in Grok Build, a mode for long-running autonomous task execution that lets a user hand off a larger implementation task with a single-line objective. Grok Build plans an approach, breaks the work into a progress checklist, fans out subagents to execute in parallel, and continues until the task is completed and verified, rather than requiring back-and-forth exchanges. Users can monitor and steer the process with companion commands (`/goal status`, `/goal pause`, `/goal resume`, `/goal clear`) and continue supplying instructions while the agent works.

Numerical and visual data

Google Trends

The graph below shows Google Trends data for global search interest in "Grok" (using the disambiguated Knowledge Graph topic, avoiding conflation with the unrelated verb "to grok"), as observed on 5 August 2026. Search volume remains minimal and flat through 2023 and most of 2024, before a sharp single-week spike to its all-time peak around March 2025, coinciding with the Grok 3 release and its associated benchmark controversy. Interest partially recedes afterward but resumes a steady overall upward trend through 2025 and into 2026, punctuated by smaller spikes plausibly tied to major releases (Grok 4, Grok 4.1, Grok 4.5) and recurring controversies over non-consensual image generation. By region, interest is concentrated most heavily in China, followed by Hong Kong, Sri Lanka, Ethiopia, and Pakistan — notable given that X and Grok are not officially available in mainland China.

Wikipedia pageviews

The graph below shows monthly pageview data for the English Wikipedia article "Grok (chatbot)," as observed on 5 August 2026. Traffic remains minimal and near-flat through 2024, before a sharp spike to an all-time peak of over 500,000 views around January 2025, coinciding with the Grok 3 release and its associated benchmark-manipulation controversy. Traffic falls sharply afterward but remains substantially elevated relative to the 2024 baseline, with recurring secondary peaks around mid-2025 and early 2026 plausibly tied to subsequent major releases (Grok 4, Grok 4.1, Grok 4.5) and the wave of controversies and regulatory actions over non-consensual image generation. Desktop and mobile-web traffic each roughly track the same pattern at lower magnitude, while desktop-spider (automated crawler) traffic shows a comparable secondary peak around the same January 2025 period, and mobile-app and mobile-web-spider traffic remain negligible throughout.

Meta information on the timeline

How the timeline was built

The initial version of the timeline was written by Sebastian Sanchez.

Check Detail construction for full timeline in timelines, Inclusion criteria for full timeline in timelines, and Representativeness of events in timelines.

Funding information for this timeline is available.

Feedback and comments

Feedback for the timeline can be provided at the following places:

• The talk page of this timeline

Timeline update strategy

This timeline should be revisited approximately monthly, given Grok's fast release cadence and xAI's frequent product announcements. Primary sources to monitor:

• xAI's official news page (x.ai/news) for model releases, product launches, and policy announcements
• Major tech outlets (TechCrunch, The Verge, Ars Technica) for independent coverage and controversy reporting
• Regulatory body announcements (EU, UK Ofcom/ICO, national telecom regulators) given Grok's recurring regulatory friction
• Elon Musk's public statements on X, treated as primary source material for future-model announcements but cross-checked against official xAI channels where possible

See also

• Timeline of OpenAI
• Timeline of ChatGPT
• Timeline of Google Gemini
• Timeline of Claude
• Timeline of DeepSeek
• Timeline of large language models
• Timeline of transformers
• Timeline of AI takeoff debates

External links

• xAI official website
• xAI news and announcements
• Grok — official chat interface
• xAI on X
• Grok on X
• GitHub — xAI open-source repositories
• xAI API documentation
• Wikipedia article on Grok (chatbot)