September 29, 2026
Less waiting in Codex: Ultrafast delivers up to 300 tokens per second
On Sep 29, OpenAI launched Ultrafast in Codex: GPT-6 Astra delivers up to 300 tokens per second, up to 8x faster than Standard. The mode is already available in Codex, ChatGPT Work, and the API. Individual users need Pro 500 for $500 per month.
Higgsfield AI 🧩
@higgsfield_ai
Breaking: @OpenAI launches Ultrafast. In Codex, it runs up to 8 times faster: 300 tokens per second. In the API: up to 6 times faster. They also mentioned "Pro 500." 👀
· 5.3K views
GPT-6 Astra previously ran on Standard in Codex. Ultrafast now speeds up token generation by up to 8x, but the measurement does not include total task execution time.
Access by price. In ChatGPT, select Ultrafast in the model picker on Pro 500. Pro 100 for $100 and Pro 200 for $200 do not include this mode. In the API, specify `service_tier: "ultrafast"` in every request for GPT-6 Astra.
The new Pro 500 offers an included limit 25 times higher than ChatGPT Plus and includes Ultrafast.
