SpaceXAI releases Grok 4.6 for complex agent tasks
Grok 4.6 is out of the rumor window. SpaceXAI says the model is built to stay with long jobs, matches GPT-5.6 Sol on the Artificial Analysis Intelligence Index, and is live in Cursor, Grok Build, and the API with a week of double included usage.
The date slipping is over. Per the August 12 launch post, SpaceXAI (formerly xAI) released Grok 4.6 with a stated focus on long-running agents and more ambitious interactive and visual work. The company says it stays with complex tasks across many steps: research, analysis, codebases, and turning an idea into a working artifact.
What SpaceXAI is claiming
The headline comparison is a tie, not a knockout. SpaceXAI says Grok 4.6 matches GPT-5.6 Sol on the Artificial Analysis Intelligence Index, a composite of nine benchmarks. On the company’s own eval table, Grok 4.6 High is listed at 61 on that index, even with Sol Max, and a point behind Fable 5 Max at 62.
The more interesting rows are the ones about staying on a job. GDPVal-AA v2 moves from 1526 on Grok 4.5 High to 1753. CursorBench v3.2 goes from 66.7% to 69.9%. DeepSWE v1.1 jumps from 54% to 65.9%, still behind Sol Max at 73%. Terminal-Bench v3.0 goes from 15.7% to 26%, still behind Sol and Fable in the mid-30s. Those are company-reported figures using “best of self-reported or public” competitor scores.
Where you can actually use it
Grok 4.6 is live in Cursor and Grok Build, plus the API and partners including OpenRouter, Vercel, and Cloudflare. SpaceXAI is offering 2x included usage in Grok Build and Cursor for the first week. API pricing starts at $2 per million input tokens and $6 per million output tokens. A fast variant is listed at twice that price.
Two days later the company also posted Grok 4.6 in GitHub Copilot. The model is not hiding in a leak screenshot anymore. It is in the tools people already pay for.
The 4.7 rumor is still a rumor
Elon Musk has already talked about a Grok 4.7 on a roughly three-to-four-week horizon, with extra SpaceX company data in supplemental training. That is an executive estimate, not a ship date. Grok 4.6 is the product that exists today.
Analysis: SpaceXAI finally put a number on the thing people were counting down to, and the number is “tied with Sol on a composite, better at not getting lost, still behind on some coding suites.” The first-week Cursor bonus is the real distribution play. Watch whether 4.6 stays selected in Cursor after the extra usage expires, and whether 4.7 lands as a model or as another calendar promise.
Watch the ByteForward breakdown: Grok 4.6 Changed the AI Race — Gemini 3.7 Flash + DeepSeek 0813 Just Landed.
Maya Chen
Related Posts
One comment
Comments are closed.
[…] 4.5 is relevant to teams comparing models for coding and agent work. For the later release, see our coverage of Grok 4.6. The two launches should be evaluated using their own dated […]