OpenAI has released a comprehensive API guide for GPT-6 Astra, walking developers through the model’s new capabilities and key migration differences. Astra supports async tool calling (the model continues reasoning while an external tool runs), mid-turn steering (instructions can be changed mid-task without restarting), and dynamic reasoning (effort level can be raised or lowered mid-dialogue while preserving prompt cache). The guide also covers computer use, structured outputs, multi-agent orchestration, programmatic tool calling, and compaction.

Migration brings several breaking changes: reasoning_effort: none is no longer supported (start with low), tool calling requires the Responses API, and temperature, top_p, and top_logprobs are unsupported. OpenAI notes that Astra often uses significantly fewer output tokens per task, so effective cost may be lower despite a higher per-token price. For autonomous agents, the company recommends explicitly instructing Astra to complete tasks end-to-end, as the model tends to ask clarifying questions more frequently.

GPT-6 Astra API Guide

Related: OpenAI Launches GPT-6 Astra with Strong Benchmark Results, The Information Reveals Astra’s Recurrent Depth Architecture