A new base model at the old price
SpaceXAI released Grok 4.7 on September 21, 2026. Unlike Grok 4.6, which reused its predecessor's base with more training, Grok 4.7 uses a new, larger base model, trained with a longer reinforcement learning run weighted toward problems that take many hours to finish. SpaceXAI says it is better at checking its own work and at managing long context. [1][5]
API pricing stayed at $2 per million input tokens and $6 per million output tokens below 200,000 prompt tokens, doubling above that, with the same 500,000-token context window. A Fast variant, the same model at twice the output speed and twice the price, is sold only through Cursor and SpaceXAI's Grok Build, not the public API. At launch Grok 4.7 was available in Cursor, Grok Build, the API and third-party routers and cloud platforms. [1][4]
Atlas interpretation: Cursor as a first launch surface, and a fast tier sold only through Cursor and Grok Build, follow from SpaceX owning Cursor since August. The coding tool is now a distribution channel for its parent's model. [4]
Two points up on the index, at twice the tokens
Artificial Analysis scored Grok 4.7 at 46 on its Intelligence Index at the xhigh reasoning setting, two points above Grok 4.6. The gains were concentrated in agentic knowledge work: 1657 Elo on its private AA-Briefcase benchmark, up 111, and 56 on its Coding Agent Index when run inside Grok Build, up nine points. Elsewhere Grok 4.7 broadly matched Grok 4.6, with small regressions on long-context reasoning. [2]
It used about 81,000 output tokens per Intelligence Index task, against about 36,000 for Grok 4.6 and 27,000 for GPT-6 Astra at maximum effort. [2]
Atlas interpretation: The unchanged list price is the headline, but the token count is the cost. A model that writes more than twice as many tokens per task at the same per-token price costs roughly twice as much to run on the same work, which narrows the price advantage over the models it still trails. [2]
Sources
- Introducing Grok 4.7
SpaceXAI · Sep 21, 2026
- Benchmarking Grok 4.7
Artificial Analysis · Sep 21, 2026
- xAI launches Grok 4.7 at bargain prices, but benchmarks reveal a wide gap to Claude and GPT-6
The Decoder · Sep 21, 2026
- Release Notes
SpaceXAI · Sep 21, 2026
- Introducing Grok 4.6
xAI · Aug 12, 2026