OpenAI has added an Ultrafast mode to GPT-6.1 Sol, and the deal is straightforward: much faster output for a lot more money. OpenAI says it runs up to 8x faster than Sol in Standard mode with intelligence close to GPT-6 Astra, while the API price is six times the Standard rate. In Codex and ChatGPT Work, only Pro $500 and eligible Enterprise and Edu plans can use it for now.
The OpenAI Developers account announced it on X at 02:26 Beijing time on October 9, saying Ultrafast is rolling out today for GPT-6.1 Sol in the API, Codex and ChatGPT Work. The 8x figure is OpenAI's own and hasn't been measured independently. Tibo (@thsottiaux) posted at 03:15 Beijing time and called it day 4 of the 28-day Codex plan. He says steering is now instant: the model reacts much faster when you adjust it, so you can change direction in real time and it doesn't waste effort. He says the two work very well together.
What it costs in the API
OpenAI's docs say Ultrafast is now available to all API users for both GPT-6.1 Sol and GPT-6 Astra. You switch it on by setting the service tier to ultrafast on gpt-6.1-sol. On OpenAI's pricing page, GPT-6.1 Sol Ultrafast costs $12 per million input tokens, $0.60 for cached input, $15 for cache writes and $60 for output at short context. That is exactly six times the Standard rates of $2, $0.10, $2.50 and $10. At long context it rises to $24 for input and $90 for output. Fast mode, the existing middle option, costs twice Standard.
Put those numbers side by side and Sol Ultrafast's input and output prices sit a little above GPT-6 Astra in Standard mode, at $10 and $50. Cached input is the exception: $0.60 against Astra's $1. So you pay for Ultrafast to save time, not money.
Ultrafast has its own rate limits, separate from Standard and Fast. For GPT-6.1 Sol the defaults are 1 million tokens per minute on the Build tier, 4 million on Launch and 40 million on Grow, well above Astra Ultrafast's 500,000, 1 million and 5 million. Sol Ultrafast also supports US and EU data residency, while Astra Ultrafast is US only. OpenAI recommends WebSockets for agent apps that make many tool calls in a row, because connection overhead can eat into the speed gain.
Who gets it in Codex and ChatGPT Work
In Codex and ChatGPT Work, Ultrafast is limited to Pro $500 and eligible Enterprise and Edu plans. OpenAI says other self-serve plans don't get it at launch, even if you've bought credits. On Pro $500 it draws on your included usage first and moves to credits once that runs out.
It also burns through usage quickly. Ultrafast counts against included subscription limits at 8x the Standard rate, and bills purchased credits and Enterprise pay-as-you-go usage at 6x. On Enterprise it's off by default until a workspace owner turns it on for selected users or the whole workspace, and older Enterprise plans that rely on rate limits instead of usage-based billing aren't supported. If you run Codex with an API key, you pay the API token prices above instead.
We're following the 28-day plan as it unfolds. Our earlier story "Tibo announces 28-day plan: daily practical Codex improvements or a full reset" covers what the team promised.
