OpenAIapi
September 29, 2026·Source

Ultrafast for GPT-6 Astra

A premium speed tier that makes GPT-6 Astra generate text up to 8x faster in Codex, for Pro $500, eligible Enterprise and Edu plans, and API developers willing to pay more per token.

Ultrafast for GPT-6 Astra announcement image
  • In Codex it generates tokens up to 8x faster than GPT-6 Astra in Standard mode, around 300 tokens per second; that is output speed, not total task time
  • In the API you set model gpt-6-astra with service_tier ultrafast, and OpenAI recommends WebSockets for agents with many quick tool calls
  • API pricing is $60 per million input tokens and $300 per million output tokens up to 272K context, versus $10 and $50 for standard Astra
  • In ChatGPT and Codex it draws down included usage at 8x the standard rate, or purchased credits at 6x
  • Available on Pro $500 and eligible Enterprise and Edu plans (off by default for Enterprise until an owner enables it); other self-serve plans can't use it even with purchased credits
  • Supports US data residency and global processing only, not EU or other regional endpoints
  • Follows August's limited Ultrafast preview on GPT-5.6 Sol, which stays in preview; GPT-6.1 Sol support is due in the coming days

Best for

Interactive agent and coding sessions where waiting on the model is the bottleneck.

When to use it

Speed matters more than cost, like live pair-programming or agents making many quick tool calls in a row.

Who benefits

Developers and heavy Codex users on the top Pro plan or an Enterprise agreement.

Join the newsletter to get ahead with AI.

I share strategies, new modules, free guides, and tools to get you better at using AI in your work.

Leading a team? See team programs →