SpaceXAI releases Grok 4.6, claiming GPT-5.6 Sol and Claude Fable 5-level intelligence

SpaceXAI has recently released an upgraded version of its Grok AI model, named Grok 4.6. The company claims that its latest release is comparable to OpenAI's GPT-5.6 Sol and Anthropic's Claude Fable 5 in terms of intelligence. Grok 4.6 is specifically designed for long-running agents, coding, knowledge work, and more complex visual projects. It is capable of maintaining focus over numerous steps, tackling unfamiliar codebases, and self-testing its output before proceeding further.

The Artificial Analysis Intelligence Index indicates that Grok 4.6 achieves a score of 61, matching GPT-5.6 Sol at its maximum reasoning level and falling short of Fable 5 Max by just one point. Additionally, SpaceXAI reports that Grok 4.6 outperforms Grok 4.5 in all evaluations, including CursorBench, FrontierCode, APEX-Agents, and Terminal-Bench.

This improvement is attributed to a longer supplemental training phase, stronger engineering data, and enhanced reinforcement learning for coding and knowledge work. Grok 4.6 is now accessible through various platforms such as Cursor, Grok Build, the SpaceXAI API, OpenRouter, Vercel, and Cloudflare. The API pricing begins at $2 per million input tokens and $6 per million output tokens, with a faster version priced at double the cost.

Users of Cursor and Grok Build enjoy twice the included Grok 4.6 usage during its initial week. SpaceXAI acquired Cursor in June, and the companies have since introduced Grok Bot for iPhone, Mac, and other devices.