SpaceXAI’s Grok 4.6 just landed on the Artificial Analysis Intelligence Index with a score of 61 — dead even with OpenAI’s GPT-5.6 Sol. Only Anthropic’s Claude Opus 5 (63) and Claude Fable 5 (62) rank higher. That’s a five-point leap over Grok 4.5, and it puts xAI squarely in the frontier conversation for the first time in a while.
The agentic performance is where things get interesting. On GDPval-AA v2, a benchmark designed to measure real-world knowledge work on a computer, Grok 4.6 posts an Elo of 1,753 — second only to Claude Opus 5. But here’s the kicker: Grok 4.6 completes complex tasks in roughly 53 steps, while Opus 5 needs about 103. Nearly half the steps suggests the model is either more decisive or more efficient at chaining actions, though benchmarks rarely tell the whole story about how that efficiency plays out in messy, real-world workflows.
Pricing is the headline-grabber. Grok 4.6 stays at $2 per million input tokens and $6 per million output tokens. That’s more than 60 percent cheaper than Claude Opus 5 at $5/$25 and GPT-5.6 Sol at $5/$30. When a model matches the best available and undercuts it by that margin, procurement teams notice. The model is live now through the API, Cursor, Grok Build, and partners including OpenRouter, Vercel, and Cloudflare. xAI is also doubling usage quotas in Grok Build and Cursor for the first week.
This is the same playbook DeepSeek used to shake up the market earlier this year — match frontier performance, slash the price, and let developers spread the word. Whether Grok 4.6 holds up outside benchmark conditions is the real question, but for teams watching their inference bills, a 60 percent discount on comparable capability is hard to ignore.
