Updated

A tall monolithic black obsidian slab with razor-sharp edges on a dark background, abstract hero artwork for Grok 4.7

xAI

Grok 4.7

xAI's newest Grok, trained on many-hour problems and built to check its own work.

Chat with Grok 4.7 in Super AgentView pricing

What is Grok 4.7?

Grok 4.7 is xAI's flagship for coding and knowledge work, released on September 21, 2026 as the successor to Grok 4.6.

xAI says it uses a new, larger base model, trained with a longer reinforcement learning run weighted toward problems that take many hours, so it works longer on hard tasks and checks its own work more carefully. It is also better at creating documents and presentations, and xAI reports that it finishes long tasks in fewer steps and output tokens than other frontier models. Its 500K-token context window takes text and images.

Provider
xAI
Context window
500K tokens
Input
Text, Image
Reasoning
Low to max
Knowledge cutoff
May 2026
Released
Sep 21, 2026

How Grok 4.7 performs

Published by xAI

Results xAI reported for Grok 4.7. Each lab runs its own setup, so compare these with other models only inside the same table.

Benchmark results xAI published for Grok 4.7
BenchmarkScoreSetting
DeepSWE v1.171%high effort
Terminal-Bench 4.038%extra high effort
SWE-Marathon v1.146%high effort
CursorBench 4.046.3%extra high effort
EEBench66%extra high effort
CADGenBench44.4%high effort
HealthBench Professional56.7%extra high effort
LatchBio Capabilities v1.044.5%extra high effort

Source: media.x.ai

What you can do with Grok 4.7

  • New, larger base model than Grok 4.6

  • Checks its own work more carefully on long tasks

  • Better than Grok 4.6 at documents and presentations

  • Fewer steps and output tokens on long tasks, according to xAI

  • A new safeguard stack that xAI rates its strongest yet against jailbreaks

How to prompt Grok 4.7

  • Ask it to verify its result before it answers; self-checking is part of how it was trained.
  • For drafts, give it the audience and the format of the final document.
  • Raise effort for long tasks. Most of xAI's published results use high or extra high.

Where Grok 4.7 fits best

  • Long tasks that benefit from self-checking

  • Drafting decks, documents and outlines

  • Working through long transcripts or notes

  • Multi-step agent runs

How Grok 4.7 compares

Grok 4.7 compared with other chat models on Flashloop
ModelLMArena textLMArena codingContext windowReasoning
Grok 4.71441#961493#91500K tokensLow to max
GPT-6 Luna1442#911507#651.05M tokensOff to max
DeepSeek V4 Flash1436#1051484#1061M tokensOff, medium, high, extra high
GPT 5.6 Luna1453#751502#761.05M tokensOff to max
Gemini 3.5 Flash Lite1455#711502#771M tokensLow, medium, high

Good to know before you start

  • Its 500K-token context is half the 1M offered by several other models in Super Agent
  • Reasoning can't be disabled
  • xAI's model card says it isn't meant for unsupervised high-stakes decisions in medicine, law or finance

Start chatting with Grok 4.7

Chat with Grok 4.7 in Super Agent

  1. Select Grok 4.7

    Open Super Agent on Flashloop and choose Grok 4.7 in the model picker.

  2. Enter your prompt

    Ask for what you need: an answer, a plan, or the images and videos you want the agent to make.

  3. Send and keep going

    Send the message. Grok 4.7 answers in Super Agent and runs any generations you asked for.

Questions about Grok 4.7

What is new in Grok 4.7?

xAI says Grok 4.7 uses a larger base model than Grok 4.6 and a longer reinforcement learning run focused on many-hour problems. It checks its work more carefully and drafts better documents and presentations.

What is Grok 4.7's context window?

500K tokens, with text and image input.

Which reasoning levels does Grok 4.7 support in Super Agent?

Super Agent's picker lists low, medium, high, extra high and max, and starts it at low. xAI's docs list low, medium, high and extra high, and describe extra high as the maximum reasoning depth.

Why do xAI's model cards say SpaceXAI?

xAI's model cards use the name SpaceXAI, which they describe as a doing-business-as name of xAI that can be used interchangeably with it. xAI's website brands the company SpaceXSI.

More chat models on Flashloop

Browse all models