Grok 5: Release Date & All We Know So Far

Update, September 23, 2026: Grok 5 is still not out. Since this guide first ran, SpaceXAI, the company xAI became when SpaceX absorbed it, has shipped Grok 4.6 on August 12, 2026 and Grok 4.7 on September 21, 2026, along with Grok Bot and Grok Voice Transcribe 2.0. None of that is Grok 5. One correction most roundups still get wrong: Grok Voice did not launch in June 2026. Voice mode has been in the Grok app since February 2025, and the developer-facing Grok Voice Agent API went live on December 17, 2025. The Q1 and Q2 2026 windows both closed empty, and the Q3 window is down to its last week. Here is the current reality check.

If you feel like the AI news cycle is spinning out of control, you aren’t alone. One minute you are mastering the latest tools, and the next, social media is exploding with rumors about a “Grok 5” that supposedly changes everything. With SpaceX officially acquiring xAI in February 2026 and a massive $20 billion Series E funding round pushing xAI’s valuation to $230 billion, the hype machine has kicked into overdrive, blending genuine technological breakthroughs with pure fan fiction. It is becoming increasingly difficult to tell if we are looking at a confirmed product launch or just another viral wishlist.

That is exactly why we wrote this guide. Instead of adding to the speculation, we are hitting the pause button to look at the hard evidence. We will walk you through exactly what xAI has officially confirmed, what shipped in the past few weeks, and which “facts” you should ignore completely. Whether you are a developer, a creator, or just curious, this is your reality check on what is actually coming next.

The key takeaways

• The Q1 2026 window Elon Musk originally targeted passed, the Q2 (May to June) window xAI later pointed to closed with no Grok 5 release, and the Q3 window is down to its last week. As of September 23, 2026 there is still no confirmed date.
• Grok V9-Medium, a 1.5 trillion parameter coding-focused model trained on Cursor data, shipped as Grok 4.5 on July 8, 2026, and SpaceXAI has stacked Grok 4.6 (August 12) and Grok 4.7 (September 21) on top of it since. All three are coding and agentic models, not Grok 5.
• Grok Voice is not new. Voice mode has been in the Grok app since February 2025 and the Grok Voice Agent API launched on December 17, 2025, so the widely repeated June 2026 launch date is wrong. The newest voice release is Grok Voice Transcribe 2.0, on September 18, 2026.
• Grok Imagine Video 1.5 (API access from June 3) took #1 on the Image-to-Video Arena when it landed, and since July it generates native 1080p from text as well as from an image, in clips of up to 15 seconds with native audio.
• The leading candidate for Grok 5 itself remains a 6 trillion parameter Mixture-of-Experts model training on Colossus 2, potentially scaling the multi-agent system from 4 to 16+ agents. SpaceXAI has still not committed to a date.

Where to verify Grok 5 news (official): the xAI News page (release posts), the official @xAI handle on X, and the Grok changelog. By bookmarking these official channels, you can avoid the wave of misinformation that typically floods social media weeks before a real launch.

What xAI shipped instead of Grok 5

While Grok 5 stayed in training, the company was anything but quiet. Releases landed in a steady run from late May through late September, and most of them get folded into “Grok 5 is here” headlines that are simply wrong. Here is what actually shipped.

Grok V9-Medium, then 4.6 and 4.7 (coding models, not Grok 5)

Elon Musk confirmed in the summer that Grok V9-Medium had finished training. It is a 1.5 trillion parameter model, roughly triple the size of the coding model it succeeds, and it was explicitly trained on Cursor data, meaning real-world developer workflows from one of the most widely used AI code editors. xAI publicly announced this model as Grok 4.5 on June 28, 2026, confirming the V9 foundation and Cursor-trained coding focus, and released it broadly on July 8, 2026 at $2 / $6 per million tokens. Two more flagships followed on the same coding-and-agents track: Grok 4.6 on August 12, 2026, and Grok 4.7 on September 21, 2026, which Artificial Analysis scores at 46.3 on Intelligence Index v4.3.2 against Grok 4.6’s 44.3, at the same $2 / $6.

This is where the naming confusion bites. V9-Medium is a separate, coding-focused foundation model. Grok 5 is the 6 trillion parameter model still training on Colossus 2. Plenty of headlines have already merged the two, so if you see “Elon unveils Grok 5 with 1.5 trillion parameters,” that is V9-Medium being mislabeled.

Grok Voice is older than the headlines say

You will see it written almost everywhere that Grok Voice launched on June 4, 2026, usually traced to a short Musk post telling people to “try Grok Voice.” That date does not survive a check against xAI’s own records. Voice mode shipped in the Grok iOS app back in February 2025 alongside Grok 3, and the Grok Voice Agent API launched on December 17, 2025. xAI’s developer release notes list no voice release at all in early June 2026; the voice model line runs Think Fast 1.0 on April 23, 2026, custom voices on May 1, and Think Fast 2.0 on July 29, 2026 at $0.08 per audio minute.

If you just want it working rather than dated, our guide to how to enable Grok voice mode covers the app steps, the six voices you can pick from, and which plans include it.

What did change in June 2026 was billing, not capability. xAI replaced the separate daily limits on each product with one shared weekly usage pool covering Chat, Imagine, Voice and Build, which is also why the per-day voice minute caps still quoted in most guides no longer exist. Voice still runs inside the Grok mobile app, so if you want Grok on a bigger screen, a desktop app is the missing piece. That is where Fello AI comes in; it puts Grok alongside ChatGPT, Claude and Gemini on your Mac in a single native app, so you are not stuck on your phone to use it.

Grok Imagine Video 1.5 took the top spot

xAI shipped Grok Imagine Video 1.5 as a preview on May 30, with API access from June 3. It turns a single still image into fluid, cinematic video with camera moves and physics, and its headline feature is native audio, generating background music, sound effects, and lip-synced dialogue in the same pass as the video. Clips run up to 15 seconds, and since July the model generates native 1080p for both text-to-video and image-to-video. It took #1 on the Image-to-Video Arena leaderboard on release, a +52 Elo jump over Imagine Video 1.0.

What we know vs what people are guessing (September 2026)

If it isn’t on xAI’s own site or reported by named journalists at major outlets, it goes in the “guessing” bucket. Here is the state of play as of September 23, 2026.

Why people think Grok 5 is coming

Even without a release, there are logical reasons the rumor mill keeps spinning.

• Rapid iteration. xAI jumped from Grok 4 to Grok 4.1 in late 2025, released Grok 4.20 Beta in February 2026 with a multi-agent architecture, shipped Grok 4.3 on April 30, and then stacked Grok 4.5 in July, Grok 4.6 on August 12 and Grok 4.7 on September 21 on top of it, with Imagine Video 1.5, Grok Bot and two voice models in between.
• Compute scale. Colossus 2 is the world’s first gigawatt-class training cluster, running seven models at once including a 10 trillion parameter giant. That kind of infrastructure does not get built for incremental updates.
• Narrative fuel. The SpaceX deal and the $20 billion funding round create massive pressure for a “big splash” launch to justify the $230 billion valuation.
• Musk’s own words. Elon Musk has publicly stated his “estimate of the probability of Grok 5 achieving AGI is now at 10% and rising,” which is bold even by his standards.

What is Grok 5

You can think of AI models like smartphone generations; they evolve rapidly, each iteration promising to be faster and smarter than the last. Grok 5 refers to the anticipated next generation of the xAI Grok language model, and the leading candidate is the 6 trillion parameter Mixture-of-Experts model training on Colossus 2.

Today the landscape runs Grok 4.7 (September 21, 2026) as SpaceXAI’s frontier model for coding and agentic work, with Grok 4.6 still behind the consumer app’s Expert and Heavy modes, and the cheaper Grok 4.3 (April 30, 2026) still available for value and long-context work, alongside Grok 4.20 Beta (February 2026), which introduced the 4-agent multi-agent architecture where specialized agents named Grok (coordinator), Harper (research), Benjamin (logic/math), and Lucas (contrarian analysis) work in parallel to cross-verify outputs. Grok 4.20 still holds a 2 million token context window and an industry-leading 83% non-hallucination rate. Grok 5 represents the next massive leap beyond all of them.

The next step for xAI Grok

Here is what to watch for in Grok 5 itself, separate from the coding and media models that shipped in the meantime.

• Expanded multi-agent system. Grok 4.20 uses 4 specialized agents. Grok 5 is expected to scale this to 16+ agents working in parallel for more complex reasoning.
• Massive scale. Colossus 2 is training models at 1T, 1.5T, 6T, and 10T parameters. The leading candidate for “Grok 5” appears to be the 6 trillion parameter Mixture-of-Experts model.
• Reliability. With 4.20 already hitting an 83% non-hallucination rate, Grok 5 is expected to push accuracy even further while tackling complex math and coding.

Note: AI naming conventions get confusing, and xAI’s internal foundation-model version numbers (like V9) do not map cleanly onto the consumer “Grok 4.x” release names. Always check official sources like x.ai before assuming a model number is final.

Grok 5 release date estimates

The biggest question is the Grok 5 release date. As of September 23, 2026, the Q1 2026 window Elon Musk originally targeted has passed, the Q2 (May to June) window xAI later pointed to has closed with no Grok 5 release, and the Q3 window has a week left on it. Three flagships landed in between: Grok 4.5 in July, Grok 4.6 in August and Grok 4.7 in September. xAI posts release notes on its News page and rolls updates across grok.com, X, and its mobile apps.

Confirmed vs rumored timelines

• High confidence. The 6T Grok 5 model is training on Colossus 2 now, and SpaceXAI has been shipping adjacent releases at a steady clip: Grok 4.5 in July, Grok 4.6 in August, and Grok 4.7, Grok Bot and Grok Voice Transcribe 2.0 through September.
• Medium confidence. Access will likely follow a tiered rollout, starting with SuperGrok Heavy ($300/month), then standard SuperGrok ($30/month) and X Premium+ ($40/month), then API customers, mirroring how xAI rolled out Grok 4.5, which reached iOS, Android, web and X two weeks after launch. Grok 4.7 went the other way and landed in Cursor, Grok Build and the API first. See our full Grok pricing breakdown.
• Low confidence. The June 30, 2026 window Polymarket tracked passed with no Grok 5 release, and the Q3 2026 window is now days from closing the same way.

Until xAI breaks its silence, treat these estimates as flexible rather than hard deadlines, and it will save you from disappointment.

Expected Grok 5 features

What will actually make this model different from what you are using today? Because xAI is connected to the X platform, Grok’s capabilities focus on the “now.”

• Real-time news. Access to breaking events as they are posted to X.
• Multimodal. With Grok Imagine already generating video with native audio, future Grok models will likely have vision and audio natively integrated within a unified architecture.
• Reasoning. With 4.20 already achieving an 83% non-hallucination rate, Grok 5 is expected to push reliability even further while rivaling specialized reasoning models in math and coding.
• Scale. The 6 trillion parameter MoE architecture means only a subset of parameters activate per query, keeping inference costs manageable despite the massive model size.

These upgrades would shift Grok from a capable multi-agent chatbot into a frontier system competing directly with GPT-6 Sol and Claude Opus 5.5 on every benchmark. Some commentators speculate xAI could add a feature dubbed the “Reality Engine,” a system to verify claims against Community Notes data in real time, though that term appears only in speculative blog content and is not an official xAI label.

Grok vs GPT and others

Choosing an AI can feel like choosing between Android and iPhone. In the Grok vs GPT debate, it usually comes down to personality versus polish. Grok offers a raw, real-time energy that fits if you want unfiltered access to live conversations, while models like Claude or Gemini often feel more polished and corporate, focusing on safety and seamless integration with your work documents. For a closer head-to-head, see our Grok vs ChatGPT comparison.

The core difference is utility versus personality. Grok doubles down on immediacy. It might not always be the best at writing a formal cover letter, but the line from Grok 4.20’s 83% non-hallucination rate to Grok 4.7’s frontier scores on agentic knowledge work shows that being fast does not have to mean being inaccurate.

Grok vs competitors (September 2026)

Tools like Fello AI let you access multiple models in one place, so you can test Grok, ChatGPT, Claude, Gemini, and DeepSeek side by side without paying for five separate subscriptions. If cost is your main question, start with whether Grok is free.

Sorting fact from fiction

The internet is full of Grok rumors, so filter information by source quality. Treat x.ai and the official @xAI handle as authoritative, verified tech journalists at outlets like The Verge or Bloomberg as high reliability, and “insider” accounts, Reddit threads, and YouTube hype videos as speculative. Applying this simple filter helps you dismiss engagement bait and focus on genuine developments.

• “That 1.5T model is Grok 5.” False. That is Grok V9-Medium, a coding model. Grok 5 is the separate 6T model.
• “Grok 5 will be free.” Unlikely for full access. Running a 6T parameter model costs millions, though older Grok models are available on grok.com with usage limits at no cost, so a limited free tier is possible.
• “It controls your computer.” False or exaggerated. Agentic features are a goal across the industry, but full autonomous control has not been released.
• “Grok 5 will achieve AGI.” Overhyped. Musk himself puts the probability at just 10%. Impressive if true, but far from a guarantee.

Trust note: Given the experimental nature of these tools, avoid uploading sensitive personal or financial data. See our guide on how to use Grok incognito mode for safer browsing.

Is Grok right for you

Ultimately, you want to know if xAI Grok will solve your specific problems, and it shines in use cases where speed and real-time data are assets. Start by defining your goal, whether that is finding the best AI models right now or a specific need like coding or creative writing. Check your budget next, since X Premium already includes Grok access and SuperGrok at $30/month opens up the full multi-agent experience. Then test the waters with the official Grok download, or learn how to use Grok on your Mac as an app to fold it into your workflow.

There is no perfect AI for everyone, so spending a few minutes with each tool is the only way to find the one that fits your workflow.

Conclusion

The hype around Grok 5 is not just about a version number; it represents a shifting landscape where real-time data and personality matter as much as raw compute. With xAI backed by SpaceX, a $230 billion valuation, and Colossus 2 training seven models at once including a 10 trillion parameter giant, the potential is undeniably massive.

The Q1 and Q2 2026 windows both came and went without Grok 5. Instead, xAI shipped Grok Imagine Video 1.5, upgraded the Grok Voice model line, and confirmed the V9-Medium coding model, none of which is Grok 5 itself. The 6 trillion parameter flagship is still in training, with full API access now looking like a Q3 2026 story. The open question is whether Grok 5 can deliver on the bolder promise of approaching AGI.

Next step: You do not have to wait for the future to upgrade your AI experience. Start by getting started with Fello AI today to run the same prompt across Grok, ChatGPT, and Claude at once, so you always get the best answer no matter which model is currently on top.

FAQ