GPT-6 Astra

GPT-6 Astra was released on 3 September 2026 as OpenAI's most capable model, aimed at work that runs from an initial request all the way to a finished result: complex reasoning, coding, computer use, research and document creation. The generation number is new, but the shape is not. Astra keeps the 1,050,000-token context window and the 128,000-token output ceiling that the GPT-5.6 tiers already had; what moved is the knowledge cutoff, now 30 April 2026, and the price, $10 per million input tokens against $5 for GPT-5.6 Sol. Long prompts are charged twice over: anything above 272,000 input tokens is billed at 2x input and cache rates and 1.5x output for the whole request, so most of that headline window costs double. Astra is also the most constrained model OpenAI has put at the top of its range. It does not accept custom temperature, top_p or log probabilities, it drops the 'none' reasoning effort level, and tool calling works only through the Responses API. Reasoning effort runs from low through medium and high to xhigh and max, and OpenAI applies asynchronous misalignment monitoring to supported agent runs, which can raise a safety alert or stop a conversation for review.

▸News

OpenAI's new generation did not add a single token of context. It costs two and a half times more

9/9/2026 · OpenAI API documentation

▸Videos

No videos yet.

▸Reviews