OpenAI's GPT-6.1 Sol nears flagship Astra at a fifth of the cost
The mid-tier model closes much of the gap to Astra on coding and computer use. It costs a fraction as much.
The facts
- Standard API pricing: $2 per million input tokens, $10 per million output tokens, $0.10 per million cached input tokens (95% below standard input, 50% below GPT-6 Sol's cached rate)
- On DeepSWE v1.1, GPT-6.1 Sol matches GPT-6 Astra at roughly one-fifth the cost and beats GPT-6 Sol's best score by 6.4 points
- On OSWorld 2.0's offline set, it comes within 2.1 points of Astra at maximum reasoning effort for about one-seventh the cost
- On Terminal-Bench Science 0.1, it costs $5.47 per task at maximum effort versus $23.21 for Opus 5.5 and $23.80 for Astra, though Astra still scores highest overall at 68.1%
- Available now in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu users, and via the API as gpt-6.1-sol; a faster 'Ultrafast' variant is coming to Codex with up to 8x faster token generation
OpenAI has released GPT-6.1 Sol, an update to its mid-tier GPT-6 Sol model. The company says it nearly matches flagship GPT-6 Astra on agentic coding, computer use and professional work. It does so at one-fifth Astra's standard input and output token prices. Cached input costs $0.10 per million tokens, which OpenAI says is 95% below standard input pricing and half of what GPT-6 Sol charged for cached tokens.
OpenAI's own benchmarks tell a consistent story: Sol closing on Astra rather than catching it. On DeepSWE v1.1, a software-engineering benchmark, GPT-6.1 Sol matches Astra's score at roughly a fifth of the cost. It beats the previous Sol's best result by 6.4 percentage points. On GDP.pdf, a professional-document benchmark, it outscores Opus 5.5 (with fallbacks) at under half the cost per task, and gets within striking distance of Astra at about a fifth the price. On OSWorld 2.0's computer-use benchmark it lands 2.1 points behind Astra, at roughly a seventh of the cost. The exception is Terminal-Bench Science 0.1, where Astra still leads outright at 68.1%. But Sol's $5.47 average cost per task undercuts Astra's $23.80 and Opus 5.5's $23.21 by more than 75%.
OpenAI also reports factuality gains. On de-identified ChatGPT conversations where users had flagged a prior model's error, GPT-6.1 Sol cut its error rate at low reasoning effort from 11.4% to 7.7%, a roughly 32% reduction. It now trails Astra's error rate by less than 2 percentage points, at low cost. On safety, the company says the model shows lower failure rates than GPT-6 Sol on respecting explicit restrictions and avoiding unauthorized actions during agentic tasks. There were no observed attempts to bypass an automated safety reviewer, matching both GPT-6 Sol and Astra on that measure, per the system card addendum.
GPT-6.1 Sol is live today in ChatGPT Work and Codex, for Plus, Pro, Business, Enterprise and Edu accounts. It is also live via the API as gpt-6.1-sol, at $2 per million input tokens and $10 per million output tokens. It is not yet available in Chat. OpenAI says an Ultrafast variant is coming to Codex within days, offering up to 8x faster token generation.
Why it matters
The pitch is straightforward. Most agentic and document workloads no longer need the flagship model to get flagship-adjacent results. Teams running high-volume coding or workflow agents should re-price against Sol before defaulting to Astra. But OpenAI's own science benchmark shows the gap doesn't close everywhere. For the hardest reasoning-heavy tasks, Astra still earns its premium.
Primary source
OpenAI — openai.comWritten from the primary source above by The Neural Net’s automated desk (claude-sonnet-5, theneuralnet/pipeline@1) without human review. How this works. Corrections: check the source first — if we got it wrong, the source link is the record.