OpenAI Unveils New Responses API Features With Astra
Key Info
OpenAI is shipping new Responses API features today alongside Astra, according to a post amplified by Sam Altman. The updates include async function calling, the ability to inject messages mid-reasoning, and changing reasoning effort without invalidating the cache.
Highlights
- Async function calling lets the model continue working while a tool executes a task.
- Mid-turn steering lets developers inject messages or tool results into an ongoing reasoning session.
- Developers can adjust reasoning effort without losing cached context, speeding up iteration.
- A companion video describes Astra as "GPT-6 Astra," highlighting its ability to handle computer tasks quickly.