September 23, 2026
Grok 4.7 closes in on coding leaders: 94% in Next.js tests
On September 22, Grok 4.7 scored 94% in the latest Next.js evals, while Opus 5.5, GPT-6 Sol, and Fable 5.1 each scored 97%. Rauch estimates that Grok costs 2-7 times less than those three models. In the xAI API, the model has a 500 000-token context window. Up to 200 000 prompt tokens, pricing is $2/$0,50/$6 per 1 million input, cached input, and output tokens. After that threshold, xAI charges $4/$1/$12.
Elon Musk
@elonmusk
Interesting. Grok 4.7 performs quite well for a relatively small model.
We ran fresh Next.js evals. The tally: ① Opus 5.5 [𝟿𝟽%] ② GPT 6 Sol [𝟿𝟽%] ③ Fable 5.1 [𝟿𝟽%] ④ Grok 4.7 [𝟿𝟺%] Notably, Grok is 2x-7x cheaper
· 4M views
In CursorBench 4.0, Grok 4.6 Extra High scored 41,4% at an average task cost of $6,10. Grok 4.7 Extra High reached 46,3% at $6,01, 70 141 tokens, and 88 steps.
Where to enable it. In Cursor, Grok 4.7 can be selected in chat or the Agent panel, and the model can also be switched with Cmd Ctrl + /. Cursor recommends it for difficult multi-step tasks, while keeping Composer for everyday coding, where speed and price matter more.
Grok 4.7's standard context in Cursor is 256 000 tokens, with a maximum of 500 000. After 256 000 input tokens, Standard costs twice as much, while Fast costs three times as much as Standard. Grok 4.7 Fast works in Cursor and Grok Build, costs twice as much as the standard version, and is not available through the public xAI API.
For API and Build. After creating XAI_API_KEY, call the model as `grok-4.7`. For a free start in Grok Build, xAI provides `curl -fsSL https://x.ai/cli/install.sh | bash`.
Artificial Analysis gave Grok 4.7 46 points at the high and xhigh levels and listed a 500 000-token context window. The gap to the three leaders in Next.js evals is 3 percentage points.
Next milestone: independent runs of Grok 4.7 on long-context coding-agent tasks.
Source
