GPT-6.1 Sol: Work, Codex, API Pricing and Benchmarks

An upgrade one week after GPT-6 Sol

OpenAI released GPT-6.1 Sol as an upgrade to GPT-6 Sol, which arrived with Luna a week earlier. On September 29 the new model became available to Plus, Pro, Business, Enterprise and Edu users in ChatGPT Work and Codex, and to developers through the API as gpt-6.1-sol. OpenAI explicitly said it was not yet available in ordinary ChatGPT Chat. [1][2]

The standard API price is $2 per million input tokens, $0.10 per million cached input tokens and $10 per million output tokens. OpenAI said an Ultrafast option for Codex would follow in the coming days, so this release does not count that option as available on launch day. [1]

What OpenAI tested

OpenAI says GPT-6.1 Sol matches GPT-6 Astra on DeepSWE v1.1 at roughly one-fifth the cost per task, and improves on GPT-6 Sol by 6.4 percentage points in its chosen comparison. On the OSWorld 2.0 offline set it reports a seven-point gain over GPT-6 Sol at maximum reasoning effort. These are OpenAI’s reported results under named test settings, not independent production measurements. [1]

For factuality, OpenAI reports that the share of responses with an error fell from 11.4 to 7.7 percent at low effort in a test built from conversations where users had already flagged a mistake. It warns that these difficult prompts are not representative of ordinary use. [1]

Capability and cost for work systems

Atlas interpretation: OpenAI positioned this model for coding, computer use and professional tasks at a lower cost than Astra. Its immediate placement in Work, Codex and the API makes the release relevant to teams running multi-step tasks, while the absence from Chat narrows the day-one audience. The comparison with Astra is the vendor’s evaluation claim. [1][2]