OpenAI launches GPT-6.1 Sol, says it nearly matches GPT-6 Astra and costs less

At its DevDay event on Tuesday, OpenAI introduced the GPT-6.1 Sol model just one week after releasing GPT-6 Sol. Although an expected release of GPT-6.1 Astra was scrapped following internal researcher safety concerns regarding higher deception levels and proceeding with tasks without user permission, OpenAI stated that GPT-6.1 Sol approaches the intelligence level of GPT-6 Astra for professional work, computer use, and agentic coding, while costing one-fifth of the standard input and output token prices.

The new model provides significant enhancements over GPT-6 Sol in programming, debugging, document comprehension, and multi-step workflows. Furthermore, OpenAI reported improved factual accuracy, noting that error rates for low reasoning effort dropped from 11.4% to 7.7%, remaining within 1.9% of GPT-6 Astra across all settings. GPT-6.1 Sol also demonstrates greater reliability in respecting user intent and safety limits, and shows improved transparency regarding its limitations.

Available immediately for Plus, Pro, Business, Enterprise, and Edu users in ChatGPT Work and Codex, the model is not yet accessible in Chat.