OpenAI's GPT-6.1 Sol brings cached input down to $0.10 per million tokens

OpenAI released GPT-6.1 Sol, a new model in its GPT-6 line, and the headline is price. The company says it gets close to GPT-6 Astra on several benchmarks while costing much less. The post doesn’t say what Astra is priced at or how big it is, so “close to Astra” rests on OpenAI’s word.

Cached input costs $0.10 per million tokens. OpenAI says that’s 95% less than standard input and 50% less than GPT-6, though the post doesn’t say whether the 50% is measured against GPT-6’s cached or standard rate. If the 95% figure holds, standard input works out to about $2 per million tokens. Caching pays off if you resend long system prompts or big chunks of RAG context on every call, because that’s where a bill tends to grow.

Compared with GPT-6 Sol, OpenAI points to better coding, computer use, and complex professional work. It also says the model is more transparent about its limitations and more reliable at following user intent and safety constraints. The post gives no numbers for any of that.

GPT-6.1 Sol is available starting today to Plus, Pro, Business, Enterprise, and Edu users in ChatGPT Work and Codex. The post doesn’t mention API access. It says developers can “build and run applications at scale,” but there’s no date or pricing table for the API in what was posted.

For now this is an announcement with a price tag and the company’s own comparisons. I’d wait for independent benchmarks and for the API listing before moving anything in production. One reply under the post counted seven new models in a few weeks. If you rewrite prompts after every release, the rewrites are where the time goes.