Grok 4.7

Grok 4.7 at a glance

Grok 4.7 is the most capable Grok model so far. xAI released it on September 21, 2026 and calls it its most powerful model for coding and knowledge work: long software tasks, documents and presentations, analysis that runs over many steps. It reasons before answering, reads images and knows about events up to May 2026, later than any other model in our chat.

• Maker: xAI, now part of SpaceX and branded SpaceXAI / SpaceXSI.
• Released: September 21, 2026, first for developers, in Cursor and Grok Build; xAI said the consumer apps would follow later, and reports from early October show it arriving there.
• Input: text and images. Output: text.
• Context window: 500,000 tokens.
• Knowledge cut-off: May 2026, per xAI's developer guide.
• In our chat: up to 10 messages a day, low reasoning effort.

What changed from Grok 4.6

Grok 4.7 is the third model in a fast summer series. Grok 4.5 came out in July 2026, trained together with the Cursor team, and Grok 4.6 followed on August 12 with a focus on long-running agents and visual work. According to xAI's announcement, Grok 4.7 is not just a tune-up: it uses a new, larger base model than Grok 4.6 and a longer reinforcement-learning run weighted toward tasks that take hours to finish. xAI says the result is better at checking its own work, managing long context, writing documents and presentations, and handling conversation inside its Grok Bot assistant.

xAI also describes a new set of safeguards, built to refuse genuinely dangerous requests while helping with legitimate security and biology work. We do not repeat the benchmark scores xAI published; they come from the maker and use high-effort settings that differ from ours.

Price stayed the same as Grok 4.6: $2 per million input tokens and $6 per million output tokens (OpenRouter list price, Oct 2026), with higher rates above 200,000 prompt tokens on xAI's API. A faster variant, Grok 4.7 Fast, exists only inside Cursor and Grok Build.

Why Grok 4.7 is limited in our chat

Grok 4.7 costs several times more per answer than Grok 4.20, and it writes longer, more thorough replies. To keep the site free for everyone, we cap it at 10 messages per visitor per day. Those 10 count toward the normal daily allowance of 30. The model menu shows how many Grok 4.7 messages you have left; when they run out, the chat switches you back to Grok 4.20 and the counter resets the next day.

We run it at low reasoning effort, so answers arrive after a short "Thinking…" pause rather than the long waits of maximum effort. You can attach an image to any message. It does not browse the web, read X or make pictures in our chat, and it sees only the recent part of a long conversation.

Best uses for your 10 messages

• Code that has to work. Reviewing a function, finding the real cause of a bug, or planning a refactor: review a Python function.
• Complex plans and documents. A project plan, a structured report or a presentation outline with clear sections: plan a shop launch.
• Explaining hard topics where accuracy and nuance matter: TCP versus UDP.
• Second opinions. Paste an answer from another model and ask Grok 4.7 to check it.
• Questions about early 2026. Its training data reaches further than Grok 4.20's or Grok 4.3's, though it still knows nothing after May 2026.

For quick everyday questions, save your Grok 4.7 messages and use Grok 4.20. For step-by-step maths, Grok 4.3 Thinking is usually enough.

Limits to keep in mind

• It can still be wrong. A stronger model makes fewer mistakes, but they tend to be harder to spot. Test code before you rely on it and check facts.
• No live information. Without search, it cannot tell you today's prices, scores or headlines.
• Not a professional adviser. xAI's own model card says it is not intended for high-stakes decisions in medicine, law or finance without human oversight.
• The daily cap. Ten messages go quickly if you use them for small talk.
• Length. Grok 4.7 tends to give thorough answers. If you want something short, say so, for example "answer in five lines".

Getting the most out of each message

• Put the whole task in one message: the goal, the material, the constraints and the format you want back.
• Paste full context. For code, include the error message and the relevant files, not just a description.
• Ask for a plan first on big jobs, then for the work section by section in follow-up messages.
• Use Grok 4.20 to prepare. Clean up notes or gather your questions with the fast model, then spend a Grok 4.7 message on the hard part.

Comparing options? See Grok 4.7 vs Grok 4.3.

Chat with Grok 4.7, free for up to 10 messages a day.

FAQ

When was Grok 4.7 released?

On September 21, 2026, on xAI's API, in Cursor and in Grok Build. It began appearing in the official Grok apps in early October.

Why only 10 Grok 4.7 messages a day?

It costs much more to run than our other models. The cap lets every visitor try it while the site stays free.

What happens when I reach the limit?

The chat switches to Grok 4.20 for the rest of the day. Your Grok 4.7 messages are back the next day.

Is Grok 4.7 the same as in the official Grok app?

It is the same model, reached through the API. The official app adds search, voice and other tools that our chat does not have.

Does Grok 4.7 know recent news?

Its knowledge goes up to May 2026. It cannot look up anything newer in our chat.