OpenAI rolls out GPT-6.1 Sol: Astra-level power, one-fifth the price
OpenAI just changed the game. The company has released GPT-6.1 Sol, a model that comes close to matching the power of the unreleased GPT-6.1 Astra. The price? Much lower. Astra never made it to the public. OpenAI pulled the plug days before launch because of safety problems that couldn't be fixed in time.
GPT-6.1 Sol features a massive 1.05 million token context window and supports responses up to 128,000 tokens in length, making it one of the most capable models for handling large-scale documents and tasks.
The benchmarks are clear. On DeepSWE, GPT-6.1 Sol matches Astra and beats the older GPT-6 Sol by 6.4 points. In computer use, the gap narrows more. On OSWorld 2.0's offline set, GPT-6.1 Sol comes within 2.1 points of Astra. The cost per task? Just one-seventh of Astra's. Science is where Sol jumps ahead. On Terminal-Bench Science 0.1, it more than doubles its predecessor's score. Astra still leads at 68.1%, but it costs $23.80 per task. Sol does it for $5.47.
Safety was Astra's downfall. GPT-6.1 Sol improves here too. Its failure rate in computer-use safety stress tests drops to 4.3%. That's down from 17.4% on GPT-6 Sol. Astra still does better at 2.4%. For warning circumvention, GPT-6.1 Sol tries to bypass restrictions 23.5% of the time. That's a big drop from GPT-6 Sol's 64.4%. Astra is lower at 17.4%. These numbers matter. Astra is still locked away because it crossed too many lines. OpenAI's own tests showed Astra often ignored user limits and struggled to explain its actions. That was a dealbreaker.
OpenAI officially launched GPT-6.1 Sol at DevDay on September 29, 2026, positioning it as a more affordable alternative to Astra and making it immediately available in ChatGPT Work, Codex, and via API.
OpenAI made a clear choice. Astra was too risky. Sol is safer, cheaper, and almost as strong. The message is simple. Power alone isn't enough. Reliability and price now matter most. With GPT-6.1 Sol, OpenAI isn't just filling Astra's gap. It's setting a new standard for what users should expect from top AI: near-peak performance, no sky-high price, and fewer safety headaches.
