At DevDay on September 29, OpenAI introduced GPT‑6.1 Sol and new developer tools. The company describes Sol as approaching Astra on a range of work tasks. That assessment comes from the vendor’s evaluations, not an independent RozumTech comparison.
Model access and API pricing
At launch, Sol is available through the API, ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu. The announcement does not mean it is already available in ordinary ChatGPT chat mode.
The model reference lists standard text-token rates per million: $2 for input, $0.10 for cached input reads and $10 for output. Requests exceeding 272,000 input tokens use higher rates for the entire request: twice the input and cache rates and 1.5 times the output rate. Other service tiers, cache writes and tools have separate pricing.
Ultrafast is a separate tier
The DevDay recap describes Ultrafast as a dedicated speed tier. Its GPT‑6.1 Sol version is still coming later. The API documentation recommends WebSockets to reduce overhead in frequent tool calls; the tier supports US data residency and global processing, but not EU regional processing endpoints.
Why run Codex in the cloud?
Codex Cloud runs tasks in isolated workspaces with a repository and dependencies, including when your computer is switched off. Reusable environment configurations help a team start with consistent settings.
Other announcements include a limited preview of Decisions API for questions with predefined answer choices, and computer-use support in Agents API.
Our practical selection criterion is the cost of an acceptable result after retries and human review. For a code change, define the tests in advance, inspect the file differences and then decide whether to accept it. That is more useful than choosing solely by generation speed. We examined the same distinction between token pricing and task cost in our Claude Sonnet 5.5 coverage.
Join the conversation
Stay on topic and respect other readers. Your first comment may appear after editorial review.