OpenAI Adds Async Tool Calling, Mid-Turn Steering to Responses API for GPT-6 Astra

What shipped

OpenAI shipped three new controls for long-running work with GPT-6 Astra in the Responses API (v1/responses), announced September 3, 2026.

What changed

• Async tool calling — the model keeps working while your application runs function or custom tools, then results return as they become available, instead of blocking the turn.
• Mid-turn steering — send additional instructions while a response is in progress over WebSockets, so the model can incorporate corrections or changing requirements mid-generation.
• Change reasoning effort mid-conversation — increase effort for hard segments or reduce it for routine follow-ups while preserving the cached prompt prefix.

Why a builder cares

These hit the two pain points of long agent runs: tool latency stalls the model, and a wrong premise means restarting. Async tool calling decouples tool work from generation, and mid-turn steering lets you correct course without burning a fresh context.