Elon Musk’s xAI releases Grok 4.6 AI: benchmark results and API pricing

xAI has officially announced the release of its flagship AI model, Grok 4.6. The new model significantly outperforms version 4.5 in intelligence and coding capabilities while maintaining the same API pricing, and is now available to developers on the xAI Build platform.

In its tests, xAI identifies the flagship GPT-5.6 Sol Max model from OpenAI and Fable 5 Max from Anthropic as the primary competitors. In evaluations focused on agentic programming, autonomous bug fixing, and complex mathematical computations, Grok 4.6 achieved the following results:

• AA Intelligence Index (general intelligence): Grok 4.6 High scored 61 points, tying with GPT-5.6 Sol Max (61) and coming close to Fable 5 Max (62);
• CursorBench v3.2 (coding quality): 69.9% successful solutions, which is higher than GPT-5.6 Sol Max (67.2%) but slightly behind Fable 5 Max (70.5%);
• FrontierCode v1.1 (complex programming): 61.3% compared to 60.6% for GPT-5.6 Sol Max and 63.6% for Fable 5 Max.

API Pricing and Availability

Pricing remains unchanged from Grok 4.5. The cost for using the base Grok 4.6 model via API is $2 per 1 million input tokens and $6 per 1 million output tokens. The accelerated Grok 4.6 Fast variant costs exactly twice as much.

The model is now available on the xAI Build service, the Cursor development environment, as well as the OpenRouter, Vercel, and Cloudflare platforms.