Grok 4.7 Released — xAI's Most Capable Model at the Same Price

xAI (published under the SpaceXAI brand) released Grok 4.7 on September 21, 2026 — its most capable model for coding and intelligent work with a 500k-token context and unchanged pricing ($2/$6 per million tokens), which scored 46 points on the independent Artificial Analysis index compared to 53 for leaders Claude Fable 5.1 and GPT-6.

Compiled by Sergey KostenchukPublished 2026-09-21Updated 2026-09-21

2026-09-21 Research OpenAI

🤖 Grok 4.7 Released — xAI's Most Capable Model at Grok 4.6's Price

Released on September 21, 2026: larger base model, longer RL training on multi-hour tasks, improved output self-verification. Price remains the same — $2/$6 per million input and output tokens, context — 500k tokens.

🌍 xAI is playing 'near-frontier at minimum price': $2/$6 compared to $4/$20 for GPT-5.6 Sol Max and $10/$50 for Fable 5.1 Max. Vendor benchmarks are better than Grok 4.6, but independent Artificial Analysis gives 46 points compared to 53 for leaders — competition in agentic coding is shifting from records to price.

👤 You can try it today: for free in Grok Build (x.ai/build), in all Cursor plans, or via the Grok API with a key from console.x.ai. Important: set prompt_cache_key (or x-grok-conv-id), otherwise the prompt cache won't work and input will be billed at full price.

Source 1: https://x.ai/news/grok-4-7 Source 2: https://docs.x.ai/developers/grok-4-7

Sources