OpenAI has introduced GPT-6.1 Sol, a fresh model built for professional workflows, complex coding, and computer use. The company states that the new offering nearly equals GPT-6 Astra across multiple rigorous tasks while offering significantly reduced API costs.
GPT-6.1 Sol Brings Near-Astra Performance
OpenAI explains, “GPT-6.1 Sol is an upgrade to GPT-6 Sol that delivers performance close to its more capable Astra model. The model focuses on agentic coding, computer use, and professional workloads that require multi-step reasoning and tool use.”
This model accommodates a 1.05-million-token context window alongside up to 128,000 output tokens. Developers can additionally access capabilities such as file search, web search, code interpreter, computer use, hosted shell, and image generation via the Responses API.
Affordability is a central highlight of the release. Standard pricing for GPT-6.1 Sol is set at USD 2 per million input tokens and USD 10 per million output tokens. By comparison, GPT-6 Astra is priced at USD 10 for input and USD 50 for output, making Sol available at one-fifth the standard token rates of Astra.
Benchmark Results Show the Gap
Assessments conducted by OpenAI indicate that GPT-6.1 Sol bridges a significant portion of the performance divide with Astra, even though it does not tie Astra on every single evaluation.
Within HealthBench Professional, Sol achieved a length-adjusted score of 64.2, trailing Astra’s 64.7. On SEC-Bench Pro, it registered 78.8%, falling short of Astra’s 85.4%. Additionally, Sol attained 35.1% on ExploitGym, whereas Astra reached 42.4%.
The debut of GPT-6.1 Sol coincides with OpenAI expanding its GPT-6 product family, presenting developers with an alternative choice that balances capability and expense. Its mix of agentic tools, a broad context window, and economical token pricing is aimed at organizations developing complex software workflows and AI agents.
Also Read: OpenAI Employees Shift Focus as AI Security Incidents Raise Concerns
