OpenAI announced gpt-6.1-sol at DevDay on September 29. It is described as delivering near GPT-6 Astra performance on coding, computer use and professional work at a lower price.
That comparison is OpenAI’s own evaluation and does not mean the two models are equal on every task.
At announcement it was available through the API and paid ChatGPT Work and Codex. It was not yet in regular Chat.
Cache writes and the whole-request long-input surcharge
Standard prices in the model docs checked on October 3 are, per million tokens, $2 input, $0.10 cache read, $2.50 cache write and $10 output. A cache write is the charge for first storing a prompt in the cache.
Above 272K input tokens, input and cache rates double and output is 1.5x for the full request, not just the excess. Regional processing adds 10%, and tools such as search carry a separate per-call fee.
The docs direct tool calling to the Responses API. Chat Completions is supported only without tool calling.
These prices and conditions reflect the docs checked on October 3 and may change.
Cost per completed task, not a single cache rate
For repeated long-document work, cache writes, expected reuse and the 272K-token threshold belong in one cost sheet. Looking only at the read price misses the expensive parts of long runs.
Moving an agent also means moving its API path. Changing only the model name on Chat Completions leaves tool calling unavailable.
The relevant Sol price is the cost of a completed workload after cache writes, long-input thresholds and the tool-calling API path are accounted for.