Grok 4.7

Maker

xAI

Released

21 Sep 2026

Open weights

No, you use it through the maker's app or paid API

License

Owned by the maker, not open

Memory (context window)

500 thousand tokens (how much it can read at once)500,000 tokens exactly

Price

$2 for every million tokens you send, $6 for every million tokens it writesA token is a small piece of a word · checked 17 Oct 2026 · price source

Official page

Official page for Grok 4.7 (x.ai)

• All-roundArtificial Analysis Intelligence Index

  46 index points Tied #5 of 17 models scored independent

  One combined score from many independent tests of how smart a model is.

  setting: xhigh · as of 17 Oct 2026 · source

• Chat & writingLMArena Text

  1444 Arena score #11 of 15 models scored independent

  When real people compare two AI answers side by side without knowing which AI wrote them, which one do they pick more often?

  setting: xHigh · as of 15 Oct 2026 · source

• Build websites & appsLMArena WebDev

  1639 Arena score #6 of 18 models scored independent

  When people ask two AIs to build the same website or app and compare the results, which one do they pick?

  setting: xHigh · as of 17 Oct 2026 · source

• CodingLiveBench 2.0 Coding

  84.5 score out of 100 Tied #7 of 15 models scored independent

  Fresh coding tasks with tests that check the answer, changed often so models cannot memorize them.

  setting: highest supported effort · as of 14 Oct 2026 · source

• ReasoningLiveBench 2.0 Reasoning

  82.7 score out of 100 #9 of 15 models scored independent

  Fresh logic and puzzle questions with one right answer, changed often so models cannot memorize them.

  setting: highest supported effort · as of 14 Oct 2026 · source

• New puzzlesARC-AGI-2

  61.4% #6 of 11 models scored independent

  New picture puzzles a model has never seen, so it has to reason instead of remember. Harder for AI than for people.

  setting: XHigh, $2.01 per task · as of 18 Oct 2026 · source

• Agent tasksTerminal-Bench 4.0

  37.6 % of tasks done #5 of 6 models scored independent

  Can an AI agent finish real jobs in a computer's command line, like setting up software or fixing a broken project?

  setting: xhigh, Grok Build · as of 18 Oct 2026 · source

• SpeedOutput speed

  93 tokens per second #9 of 16 models scored independent

  How many tokens (pieces of words) per second the model writes once it starts answering. Higher is faster.

  setting: xhigh · as of 17 Oct 2026 · source

• XiaomiMiMo-V2.6-ProOpen weights. License: MIT: free for almost any use46 index pointsTied #5 of 17 models scored
• Z.aiGLM-5.3Open weights. License: GLM-5.3 license: free to use with conditions45 index points#7 of 17 models scored
• Moonshot AIKimi K3Open weights. License: Kimi K3 license: free to use with conditions44 index points#8 of 17 models scored

No stories about Grok 4.7 on ITG Now yet.