OpenAI launches Ultrafast tier, boosting GPT-6 Astra token speed up to 8x

At OpenAI's 2026 Developer Day event in the US, OpenAI introduced a new service tier called Ultrafast. It can deliver up to 8x faster token generation in Codex and up to 6x faster generation through API calls.

According to official information, GPT-6 Astra Ultrafast can generate 300 tokens per second in Codex. API access is now open. Pro 500 subscribers and enterprise users can try it in ChatGPT Work and Codex. OpenAI says GPT-6.1 Sol Ultrafast will follow later.

For API pricing, gpt-6-astra short-context input costs $60 per 1M tokens and output costs $300. Long-context input is $120 and output is $450. Cached input is cheaper at $6 per 1M tokens.

The main value of Ultrafast is combining stronger model capability with real-time response speed. It suits use cases like incident response, real-time customer service, trading news generation, and interactive coding.