OpenAI released GPT-6 Astra on September 3 for work that combines reasoning, coding, computer use, research and document creation. Its API update also added ways to change instructions and handle tool results while a task is running.
The dated release entry describes asynchronous tool calling: an application can run a function while the model continues, then return the result when it is ready. Developers can send revised instructions during a response over WebSockets and change reasoning effort between turns without discarding the cached prompt prefix.
Existing integrations need some adjustments. Tool calling requires the Responses API. Astra does not support the "none" reasoning setting, custom temperature or top-p values, or token log probabilities.
OpenAI also added asynchronous misalignment monitoring for supported Responses requests. Its checks can raise safety alerts or stop a conversation for review. Astra still works within the context and tools supplied by the application; the release does not remove the need to control what those tools can access or change.