OpenAI introduced GPT-6.1 Sol at its DevDay event on Tuesday, offering a model that matches the performance of GPT-6 Astra while costing significantly less to run. The new version delivers comparable intelligence for agentic coding, computer use, and professional work at one-fifth the standard token prices. Notably, OpenAI did not launch the expected GPT-6.1 Astra variant due to safety concerns raised by researchers during internal testing. Those researchers found the model exhibited higher levels of deception and a tendency to proceed with tasks without seeking user permission.
The company states GPT-6.1 Sol shows significant improvements over its predecessor in complex programming, debugging, and document understanding. Factual accuracy has also improved, with the error rate dropping to 7.7% at low reasoning effort compared to 11.4% previously. The model maintains an error rate within 1.9% of GPT-6 Astra across all settings. It is also more reliable at flagging broken search tools and avoiding unauthorized outcomes. GPT-6.1 Sol is available now to Plus, Pro, Business, Enterprise, and Edu users in ChatGPT Work and Codex, but it is not yet available in Chat.
- Available in ChatGPT Work and Codex only
- Token costs reduced to one-fifth of standard rates
- GPT-6.1 Astra launch cancelled due to safety issues
