Summary
On August 12, 2026 xAI (SpaceXAI) released Grok 4.6, its flagship model tuned for long-running agents and multi-step interactive work. It posts a 61 on the Artificial Analysis Intelligence Index — tying GPT-5.6 Sol Max and one point behind Fable 5 Max — carries a 500,000-token context window, and lists at $2/$6 per million input/output tokens (roughly half rival frontier pricing), available via the xAI API, Grok Build, Cursor, OpenRouter, Vercel, and Cloudflare.
What changed
xAI shipped Grok 4.6, which beats Grok 4.5 High across benchmarks, adds more self-checking on longer tasks, offers a 500K-token context, and is distributed on day one through the xAI API, Grok Build, Cursor, OpenRouter, Vercel AI Gateway, and Cloudflare.
Why it matters
Frontier-competitive quality at roughly half the price, aimed explicitly at long-running agents with self-verification, pressures incumbent model pricing and strengthens the case for agent builders to route to Grok on cost. Broad same-day gateway availability shows distribution, not just capability, is now the competitive battleground.
Evidence excerpt
Grok 4.6 posts a 61 on the Artificial Analysis Intelligence Index (tying GPT-5.6 Sol Max, one behind Fable 5 Max), has a 500K-token context, costs $2/$6 per 1M input/output below 200K tokens, and is available via xAI API, Grok Build, Cursor, OpenRouter, Vercel, and Cloudflare.