Updated
Updated · xAI · Aug 12
xAI Releases Grok 4.6 at $2 per Million Tokens, Matching GPT-5.6 Sol on 9 Benchmarks
Updated
Updated · xAI · Aug 12

xAI Releases Grok 4.6 at $2 per Million Tokens, Matching GPT-5.6 Sol on 9 Benchmarks

3 articles · Updated · xAI · Aug 12

Summary

  • Grok 4.6 is now live in Cursor, Grok Build, the API, and partners including OpenRouter, Vercel and Cloudflare, with pricing from $2 per million input tokens and $6 per million output tokens.
  • The model targets long-running agents, coding and visual product work, with xAI saying it can sustain multi-step research, work across codebases and turn broad ideas into working first versions.
  • On benchmarks, xAI said Grok 4.6 matches GPT-5.6 Sol on the Artificial Analysis Intelligence Index, a composite of 9 tests, and improves on Grok 4.5 in first-pass interactive and visual output.
  • xAI attributed the gains to a longer training run, regenerated supervised fine-tuning traces, broader agentic reinforcement-learning tasks and expanded safety testing calibrated to the model’s stronger capabilities.
  • A fast variant costs 2x more, while Grok Build and Cursor users get 2x included usage for the first week as xAI pushes adoption of its latest model.

Insights

Could Grok 4.6's integration with autonomous payment wallets lead to AI agents secretly funding their own unapproved operations?
As AI models match GPT-5.6, what hidden dangers lurk in letting long-running agents execute untrusted code in live environments?
Will the shift from writing code to verifying AI-generated code actually save developers time or just create new bottlenecks?