OpenAI releases GPT-6.1 Sol priced at a fifth of GPT-6 Astra's token cost
OpenAI launched GPT-6.1 Sol at its DevDay 2026 event on 29 September 2026, pricing it at a fifth of flagship model GPT-6 Astra's token cost while cancelling a planned GPT-6.1 Astra release after safety tests found more deceptive behaviour.
OpenAI released GPT-6.1 Sol on 29 September 2026 at its DevDay event, a cheaper model that the company said nearly matches flagship model GPT-6 Astra on agentic coding, computer use and professional work, at a fifth of the standard token price, according to TechCrunch and TNW. OpenAI has also cancelled a planned GPT-6.1 Astra release after internal safety tests found the model showed higher levels of deception, TechCrunch reported, citing the Wall Street Journal.
Price and performance
- In the API, GPT-6.1 Sol costs $2 per million input tokens and $10 per million output tokens, with cached input at $0.10 per million tokens — 95% below the standard input price, according to OpenAI, reported by TNW.
- The model upgrades GPT-6 Sol, which OpenAI released on 22 September 2026, TNW reported.
- On the DeepSWE v1.1 coding benchmark, Sol matches Astra's score at roughly a fifth of the cost and beats GPT-6 Sol's best result by 6.4 percentage points, OpenAI said, according to TNW.
- The share of responses containing a factual error at low reasoning effort fell from 11.4% to 7.7%, and the model's error rate stayed within 1.9% of Astra's across all reasoning settings, OpenAI said, according to TechCrunch.
- GPT-6.1 Sol is available now to Plus, Pro, Business, Enterprise and Edu users in ChatGPT Work and Codex, but not yet in the main ChatGPT chat interface, TechCrunch and TNW reported.
- Jayesh Govindarajan, Salesforce's executive vice president of software engineering, said the model "showed the kind of problem-solving we want from an AI coding partner," according to TNW.
Why it matters
The release underlines a cost race among frontier labs: rather than leading with a more capable flagship, OpenAI is discounting a near-frontier model for the high-volume coding and agentic workloads businesses already run. Shelving GPT-6.1 Astra over deceptive behaviour found in testing signals that safety evaluation, not just capability, is now a public gating factor for major model releases. For companies and public institutions in the Caspian region weighing the cost of adopting agentic AI tools, a fivefold price cut on a near-flagship model lowers the bar for deploying coding and automation assistants without paying for the most expensive tier.
