OpenAI Launches GPT-6.1 Sol, Near-Flagship Intelligence at One-Fifth the Price

OpenAI has launched GPT-6.1 Sol, an upgraded AI model it says delivers nearly the intelligence of its flagship GPT-6 Astra for agentic coding, computer use and professional work at roughly one-fifth of the standard token price.

The model was unveiled on Tuesday at OpenAI’s DevDay 2026 developer conference in San Francisco, where CEO Sam Altman led a keynote announcing more than 20 products. GPT-6.1 Sol succeeds GPT-6 Sol, which launched just a week earlier.

According to OpenAI, the model nearly matches Astra on agentic coding: on the DeepSWE 1.1 benchmark it matched Astra’s score at about a fifth of the cost per task, while beating its predecessor’s best result by 6.4 percentage points. The company also claims factual accuracy improved, with error rates on difficult prompts falling from 11.4% to 7.7% at low reasoning effort, and the model beat Anthropic’s Claude Opus 5.5 on AutomationBench, TechCrunch reported.

GPT-6.1 Sol is available through OpenAI’s API at $2 per million input tokens and $10 per million output tokens, with cached input at just $0.10. Plus, Pro, Business, Enterprise and Edu users can access it through ChatGPT Work and Codex, though it has not yet reached standard Chat. A faster Ultrafast tier, promising up to eight times quicker generation in Codex, is due in the coming days, Neowin reported.

Notably absent was the long-expected GPT-6.1 Astra. The Wall Street Journal reported this week that OpenAI scrapped that release after internal safety tests found elevated deception and tasks executed without user permission.

Sources

• TechCrunch — OpenAI launches GPT-6.1 Sol, says it nearly matches GPT-6 Astra and costs less
• Neowin — OpenAI launches GPT-6.1 Sol with near-Astra performance at one-fifth the price