OpenAI rolls out Ultrafast mode for GPT-6.1 Sol
- Source
- x.com
- Date
Ultrafast first arrived for GPT-6 Astra on September 30. This rollout extends it to GPT-6.1 Sol, the cheaper model OpenAI released the same week at $2 input and $10 output per million tokens. The fast mode costs six times those rates, so it buys latency rather than savings. In Codex and ChatGPT Work, access requires Pro 500 or eligible Enterprise and Edu plans, while API developers can use it directly, with US and EU data residency supported. The tradeoff is clearest for interactive coding, where waiting on a model breaks flow, and for agent steps on a critical path. Batch work and background agents gain little from the speed and pay the full premium. Teams already on Sol can test Ultrafast on latency-sensitive calls only and keep standard mode everywhere else.
Ultrafast is rolling out today for GPT-6.1 Sol in the API, Codex, and ChatGPT Work. Near-Astra intelligence at up to 8x faster speeds than Sol Standard, so you can build as fast as the ideas come.
Context
OpenAIDevs announced in a post on X that Ultrafast, a faster speed mode, is rolling out for GPT-6.1 Sol in the , Codex and ChatGPT Work, describing it as near-Astra intelligence at up to 8x the speed of Sol Standard. Astra is OpenAI's GPT-6 Astra , and "near-Astra" repeats how OpenAI described Sol when GPT-6.1 Sol launched. The new claim is that Sol keeps that capability in Ultrafast mode while generating much faster.
The post follows through on an earlier one. When the same account introduced Ultrafast for Astra, it said Ultrafast ran up to 8x faster than Astra Standard in Codex and that Ultrafast for GPT-6.1 Sol was coming soon. Both speed figures are OpenAI's own claims, each measured against that model's standard mode. The post does not say how closeness to Astra was measured, its attached video could not be reviewed, and only the opening post was collected, so what Sol Ultrafast costs and which plans get it beyond the named products are not covered here.
Checking sign-in…
Loading comments…







