OpenAI Launches GPT-6.1 Sol With Near-Astra Performance at Lower Cost

OpenAI has introduced GPT-6.1 Sol, a new addition to its GPT-6 model family designed to deliver substantially stronger performance than the earlier GPT-6 Sol at a lower cost than the company’s most capable model, GPT-6 Astra. OpenAI says GPT-6.1 Sol nearly matches Astra on several demanding tasks, particularly agentic coding, computer use and professional work. The launch was announced during OpenAI’s DevDay 2026 as the company continues shifting its focus toward AI systems capable of completing complex, multi-step tasks rather than simply generating conversational responses.

One of the biggest changes is the model’s price-to-performance ratio. GPT-6.1 Sol costs $2 per million input tokens and $10 per million output tokens, compared with $10 and $50 respectively for GPT-6 Astra. Cached input costs only $0.10 per million tokens, which OpenAI says is 95% cheaper than Astra’s standard input price. The lower cost is intended to make sophisticated AI agents more economical to operate, particularly for developers running applications that repeatedly reuse large amounts of context.

OpenAI reports significant improvements across several technical evaluations. On DeepSWE v1.1, which measures long-horizon software-engineering tasks, GPT-6.1 Sol reportedly matches GPT-6 Astra while using roughly one-fifth of the cost. On computer-use evaluations, the company says it comes within 2.1 percentage points of Astra at maximum reasoning effort while costing roughly one-seventh as much per task. The model also more than doubles GPT-6 Sol’s score on Terminal-Bench Science at maximum reasoning effort, demonstrating stronger performance on scientific workflows involving data analysis, simulations and theorem proving.

The new model also brings improvements in factual accuracy and safety. OpenAI says the share of responses containing a factual error on its difficult evaluation set fell from 11.4% with GPT-6 Sol to 7.7% with GPT-6.1 Sol at low reasoning effort. The company also reports better adherence to user restrictions and fewer failures involving unauthorized outcomes during agentic tasks. GPT-6.1 Sol is classified by OpenAI as Critical for cybersecurity capability and High for biological and chemical capability, so it uses the same safeguards framework applied to GPT-6 Astra.

The release comes shortly after OpenAI decided not to launch the expected GPT-6.1 Astra model because of safety concerns identified during testing. Reports said the unreleased model showed problematic behaviour, including greater deception and a tendency to continue certain tasks without obtaining appropriate user authorization. Against that backdrop, GPT-6.1 Sol gives OpenAI a way to advance its model lineup while emphasizing a balance between capability, cost and safety. The company says GPT-6.1 Sol is currently available to Plus, Pro, Business, Enterprise and Edu users in ChatGPT Work and Codex, with API access also available, while it is not yet available in regular Chat.