OpenAI rolls out Sol Ultrafast with a sixfold API price premium

OpenAI began rolling out GPT-6.1 Sol Ultrafast on October 8 across its API, Codex and ChatGPT Work; its original model announcement put the speed claim at up to eight times faster token generation in Codex than Sol Standard. Developers can access the faster service through the API, while access inside the two work products is restricted to eligible plans.

The price changes the calculation. Ultrafast costs $12 per million input tokens and $60 per million output tokens, six times Sol’s standard rates of $2 and $10, respectively, according to OpenAI’s current pricing table for short context requests. Tokens are the units used to meter model input and output. This is a premium speed setting for the existing model, rather than a newly announced intelligence upgrade.

36Kr’s report highlights the subscription barrier: Codex and ChatGPT Work access is limited to Pro 500, eligible usage based Enterprise and credit based Edu plans, with Enterprise administrators required to enable it. Those plan names and the administrator requirement also appear in OpenAI’s October 8 rollout announcement. OpenAI’s API documentation sets a different rule: all API users can access Sol Ultrafast, subject to separate rate limits. Having access to ordinary Sol therefore doesn’t automatically give a subscriber the faster mode.

The speed claim also needs a careful reading. The October 8 rollout announcement advertises speeds up to eight times faster than Sol Standard; OpenAI’s original Sol announcement describes the eightfold improvement as faster token generation in Codex, rather than a promise that every task finishes eight times sooner. Its API guidance warns that network overhead can erode the gains and recommends persistent WebSocket connections, which keep a connection open across requests.

The next useful test is completed work per dollar. Developers need to measure whether shorter waits on their own tasks justify paying six times the standard token rates, especially when network overhead can reduce the speed gains.