Summary
OpenAI released GPT-6 Astra on September 3, 2026, its new frontier general-purpose model aimed at long-horizon, multi-step agentic work. It is rolling out to ChatGPT Plus, Pro, Business, and Enterprise and is callable through the OpenAI API as gpt-6-astra, with distribution via Amazon Bedrock and Microsoft Foundry.
What changed
OpenAI shipped GPT-6 Astra, first as a limited preview for trusted organizations on September 3 and broadening over the following days. API pricing is $10 per million input tokens and $50 per million output, with cached input at $1, batch at half price, and a 2x Fast mode. The model is positioned for autonomous coding, computer use, research, and multi-step professional workflows.
Why it matters
GPT-6 Astra is OpenAI's bid to lead on reliable long-horizon agentic execution, arriving the same week Anthropic shipped Claude Fable 5.1 at identical base token pricing. For teams building coding and agent products, it widens frontier-model choice across the OpenAI API, Bedrock, and Foundry and immediately flows into downstream surfaces like GitHub Copilot.
Evidence excerpt
OpenAI says GPT-6 Astra 'is faster and capable of performing more tasks than any prior iteration'; it is rolling out to ChatGPT Plus/Pro/Business/Enterprise and through the OpenAI API and AWS, callable as gpt-6-astra, and available on Amazon Bedrock and Microsoft Foundry. API pricing: $10/M input, $50/M output, $1/M cached input.