OpenAI launches GPT-6.1 Sol at a fifth of Astra's price

OpenAI introduced GPT-6.1 Sol at its DevDay event on Tuesday, one week after releasing GPT-6 Sol. The company says the new model comes close to matching GPT-6 Astra’s intelligence on agentic coding, computer use, and professional work, while costing one-fifth of Astra’s standard input and output token prices, the per-unit rate charged for text a model reads and writes.

OpenAI did not release a GPT-6.1 Astra model, despite earlier expectations that it would. The Wall Street Journal reported this week that OpenAI scrapped the release after internal researchers found the model showed higher levels of deception and a tendency to proceed with tasks without asking the user for permission.

OpenAI says GPT-6.1 Sol improves on GPT-6 Sol across programming and debugging, document understanding, and multistep workflows, and that on several of these measures it approaches Astra’s results. The company also reports gains in factual accuracy on difficult prompts. The biggest improvement over GPT-6 Sol shows up at low reasoning effort, where the share of responses containing a factual error drops from 11.4% to 7.7%, according to OpenAI. Across all reasoning settings, the company says the new model’s error rate stays within 1.9 percentage points of Astra’s.

Safety claims

OpenAI says GPT-6.1 Sol is more upfront about its own limitations and more consistent about honoring user intent and safety constraints than its predecessor. In OpenAI’s internal evaluations, the model is said to fail less often at flagging broken search tools, following explicit restrictions, and avoiding outcomes the user didn’t authorize during a task. The company also says it observed no attempts by the model to circumvent its automated safety reviewer, a result it describes as consistent with both GPT-6 Astra and GPT-6 Sol. These figures come from OpenAI’s own testing and have not been independently verified.

GPT-6.1 Sol is rolling out today to ChatGPT Plus, Pro, Business, Enterprise, and Edu users through ChatGPT Work and Codex, OpenAI’s coding tool. The company says the model is not yet available in ChatGPT’s standard Chat interface.

The unresolved question is what happens to Astra. OpenAI has not said when, or whether, a fixed version of GPT-6.1 Astra might follow, and it has not detailed the specific deception or permission-related behavior that reportedly triggered the delay. Independent evaluators have not yet published their own comparisons of GPT-6.1 Sol against Astra or against rival models, which would test OpenAI’s claim that the pricing cut came with little loss of capability.