GPT-6.1 Sol is out at Sonnet 5.5's per-token price, with cached input at half Anthropic's rate. For agent loops that re-read context, it goes on your eval this week.
OpenAI released gpt-6.1-sol at DevDay on 29 September. API list price is $2 per million input tokens, $0.10 cached input and $10 output.1 It is live in Codex and ChatGPT Work for Plus, Pro, Business, Enterprise and Edu, and not yet in ChatGPT's Chat.1 GitHub made it generally available in Copilot the same day for Pro+, Max, Business and Enterprise, billed at provider list price, with a gradual rollout.2
Claude Sonnet 5.5, released a day earlier, lists at the same $2 and $10, with cache reads at $0.20.3 On per-token price the two are identical except for cached input, where Sol costs half. Agent sessions are dominated by cached reads, so that line can decide the bill; how much depends on your cache-hit rate, which only your own logs can tell you.
Run it against your own task set before you move traffic. The price is verified; the capability claims are OpenAI's own runs.
What OpenAI claims, and what else DevDay shipped
OpenAI says Sol matches GPT-6 Astra on DeepSWE v1.1 at roughly a fifth of the cost, and scores 2.2 points above Opus 5.5 on AutomationBench at medium effort.1 Both are vendor measurements and we have not seen them reproduced.
Also live from the same event: GPT-6 Astra Ultrafast in the API and in Codex on Pro 500 and Enterprise; Codex in the cloud with reusable environments; the Agents API with computer use.4