SpaceXAI has launched Grok 4.6, a large language model that scores 61 points on the Artificial Analysis Intelligence Index, matching the performance of OpenAI’s GPT-5.6 Sol. This result follows a five-point improvement from the previous Grok 4.5 iteration and places the model behind only Anthropic’s Claude Opus 5 and Claude Fable 5. The new system demonstrates particular strength in agentic tasks, where it independently executes multi-step workflows. On the GDPval-AA v2 benchmark, Grok 4.6 ranks second with an Elo score of 1,753, trailing only Claude Opus 5. It completes complex computer-based tasks in approximately 53 steps, whereas Claude Opus 5 requires roughly 103 steps to achieve similar results.
The model maintains a pricing structure of $2 or $6 per million tokens depending on context length, which is more than 60 percent cheaper than competing top-tier options. This cost advantage applies to both Claude Opus 5 and GPT-5.6 Sol, which charge $5 and $30 respectively for equivalent usage. Grok 4.6 is currently accessible via the API, Cursor, Grok Build, and partner platforms including OpenRouter, Vercel, and Cloudflare. For the first week of availability, x.ai is providing double the usage quota within Grok Build and Cursor to encourage early adoption.
* Available immediately through multiple API partners
* Costs less than half of leading competitors
* Requires fewer steps to complete complex workflows
