Grok 4.6

Grok 4.6 shipped on August 7, 2026 on the same 1.5 trillion parameter V9 foundation as Grok 4.5, with the entire gain coming from improved supervised fine-tuning and reinforcement learning rather than a larger model. The context window is 500,000 tokens. Pricing has a wrinkle worth budgeting for: prompts below 200,000 tokens bill at $2 per million input, $0.50 per million cached input, and $6 per million output, but once a request crosses 200,000 tokens those rates double to $4, $1, and $12, and the higher tier applies to every token in that request rather than only the overage. Do not extrapolate the $2/$6 headline across the full 500K window when you model cost of ownership. On capability, Artificial Analysis places Grok 4.6 above Kimi K3 and level with GPT-5.6 Sol, which puts it third overall on that index at a fraction of Sol's $5/$30 rate. xAI positioned it directly against Kimi K3, at roughly 2.8 trillion parameters, and Claude Opus 4.8, and the post-training-only approach is the argument: xAI is betting that a well-tuned 1.5T model beats a larger one at the same latency and cost envelope. A larger 2.1 trillion parameter Grok 4.7 has been signaled to follow within weeks, so treat 4.6 as a checkpoint rather than a plateau. Available through the xAI API, Grok in X, and SuperGrok.