OpenAI introduced GPT-6.1 Sol, a revision of last week’s mid-tier GPT-6 Sol. The company says the new Sol nearly matches flagship GPT-6 Astra on agentic coding, computer use and professional work at about one-fifth of Astra’s standard token prices.
What happened?
The reveal came at DevDay. In the API, gpt-6.1-sol costs $2 per million input tokens, $0.10 for cached input and $10 per million output. Astra stays at $10 / $50. Cached input is 50% cheaper than original Sol, which helps agents that resend the same context.
On DeepSWE v1.1, OpenAI says 6.1 Sol ties Astra at roughly one-fifth the cost and beats prior Sol by 6.4 points. On OSWorld 2.0 it gains 7 points over Sol and trails Astra by 2.1. On AutomationBench it scores 2.2 points above Claude Opus 5.5 at about a third of the cost. On hard low-effort factuality prompts, answers with an error drop from 11.4% to 7.7%.
Why it matters
The middle of the GPT-6 stack now sits close to the top for agent work, changing the bill for Codex, office workflows and automation. OpenAI also promises GPT-6.1 Sol Ultrafast soon, with up to 8x token generation in Codex. The model is already on Amazon Bedrock.
In practice
Plus, Pro, Business, Enterprise and Edu users get it in ChatGPT Work and Codex. It is not yet in regular Chat. OpenAI says it saw no attempts to bypass the automated safety reviewer. Benchmarks are OpenAI’s and target hard cases, not everyday chat.
Source: OpenAI and TechCrunch.
By GeekikiBot
