Collect Video Testimonials That Build Trust
GPT-6.1 Sol
GPT-6.1 Sol is OpenAI's reasoning model for coding, professional document analysis, computer use, scientific research and AI agents, available through the OpenAI API, ChatGPT Work and Codex.
About GPT-6.1 Sol
GPT-6.1 Sol is a reasoning model developed by OpenAI for software engineering, professional knowledge work, computer interaction and scientific research. It supports complex tasks involving code generation, debugging, document understanding and multistep workflow execution. Developers can access the model through the OpenAI API using the identifier gpt-6.1-sol.
For software engineering, GPT-6.1 Sol handles complex programming assignments involving real codebases. OpenAI evaluates its coding capabilities using DeepSWE v1.1, which tests AI agents on original, long horizon software engineering tasks. The model is also designed to analyze complex professional PDF documents containing tables, charts, diagrams and detailed information.
GPT-6.1 Sol supports agentic applications that interact with computer environments and execute multistep business workflows. Its documented evaluations cover computer use, business process automation and scientific research tasks involving data analysis, simulations and mathematical reasoning. These capabilities make it relevant to developers building AI agents for software development, business operations and research applications.
OpenAI provides GPT-6.1 Sol through its API and makes it available to eligible subscribers within ChatGPT Work and Codex. The official announcement documents improvements in coding, professional work, computer use, scientific research and factual accuracy. The announcement does not specify the model's context window, maximum output token limit, knowledge cutoff or underlying architecture.
Key Features
- Complex Reasoning: Supports reasoning across software engineering, professional knowledge work, computer interaction, business automation and scientific research.
- Software Engineering: Handles code generation, debugging and complex programming tasks involving real codebases. OpenAI evaluates its coding capabilities using DeepSWE v1.1, which tests long horizon software engineering assignments.
- Professional Document Analysis: Understands complex PDF documents containing tables, charts, diagrams and detailed information. Its document analysis capabilities are evaluated using GDP.pdf across professional domains.
- Computer Use and Business Automation: Supports agentic workflows involving interactions with computer applications and the execution of multistep business processes. OpenAI evaluates these capabilities using OSWorld 2.0 and AutomationBench.
- Scientific Research: Supports scientific workflows involving data analysis, simulations, model fitting and mathematical reasoning using code and terminal tools. Its scientific capabilities are evaluated using Terminal-Bench Science 0.1.
- API Access and Factual Accuracy: Available through the OpenAI API for developers building AI applications. OpenAI documents improved factual accuracy on deliberately difficult evaluation prompts and reports results across multiple reasoning effort settings.
Pricing
-
Standard API Input Tokens: $2.00 per 1 million input tokens Standard input token pricing for GPT-6.1 Sol through the OpenAI API.
-
Cached API Input Tokens: $0.10 per 1 million cached input tokens Reduced input token pricing when cached context is reused across API requests.
-
Standard API Output Tokens: $10.00 per 1 million output tokens Standard output token pricing for GPT-6.1 Sol through the OpenAI API.
Use Cases
- Software engineering, code generation, debugging and complex programming assignments involving real codebases.
- Professional document analysis involving complex PDF documents, tables, charts, diagrams and detailed information.
- Building AI agents that interact with computer applications and execute multistep business workflows.
- Scientific research workflows involving data analysis, simulations, model fitting and mathematical reasoning using code and terminal tools.
Pros & Cons
Pros:
- Demonstrated software engineering capabilities on DeepSWE v1.1, including complex programming assignments involving real codebases.
- Evaluated capabilities across professional document analysis, multistep business workflows, computer use and scientific research.
- Available through the OpenAI API, with documented improvements in factual accuracy on challenging evaluation prompts.
Cons:
- Not yet available in standard ChatGPT Chat at the time of the official announcement.
- Access through ChatGPT Work and Codex is limited to eligible Plus, Pro, Business, Enterprise and Edu users.
Integrations
OpenAI API, ChatGPT Work, Codex
FAQ
What is GPT-6.1 Sol?
GPT-6.1 Sol is a reasoning model developed by OpenAI for software engineering, professional document analysis, computer use, business workflow execution and scientific research. It is available to developers through the OpenAI API.
What are the main capabilities of GPT-6.1 Sol?
Its documented capabilities include writing and debugging code, analyzing complex professional PDF documents, completing multistep business workflows, interacting with computer applications and supporting scientific research using code and terminal tools.
Does GPT-6.1 Sol support software engineering?
Yes. OpenAI evaluates GPT-6.1 Sol using DeepSWE v1.1, which tests AI agents on original, long horizon software engineering tasks involving real codebases. The model supports writing and debugging code.
Can GPT-6.1 Sol analyze PDF documents?
Yes. OpenAI evaluates its professional document understanding using GDP.pdf, a benchmark involving complex PDF documents containing tables, charts, diagrams and detailed information from professional workflows.
Can GPT-6.1 Sol execute business workflows?
Yes. GPT-6.1 Sol is evaluated on AutomationBench, which tests whether AI agents can complete multistep business workflows. The benchmark covers 47 tools across sales, marketing, operations, support, finance and human resources.
Does GPT-6.1 Sol support computer use?
Yes. OpenAI documents its ability to handle tasks requiring interactions with computer applications. Its computer use capabilities are evaluated using the offline set of OSWorld 2.0.
Can GPT-6.1 Sol assist with scientific research?
Yes. OpenAI evaluates the model using Terminal-Bench Science 0.1, which covers scientific workflows involving data analysis, simulations, model fitting and mathematical tasks using code and terminal tools.
Is GPT-6.1 Sol available through an API?
Yes. Developers can access GPT-6.1 Sol through the OpenAI API using the model identifier gpt-6.1-sol.
Where is GPT-6.1 Sol available?
According to OpenAI's official announcement, GPT-6.1 Sol is available through the OpenAI API and to eligible Plus, Pro, Business, Enterprise and Edu users within ChatGPT Work and Codex. It was not yet available in standard ChatGPT Chat at the time of the announcement.
How accurate is GPT-6.1 Sol?
OpenAI evaluates factual accuracy using deliberately difficult conversations in which users previously identified factual errors. At low reasoning effort, 7.7% of evaluated GPT-6.1 Sol responses contained at least one factual error. OpenAI notes that these evaluation prompts are not representative of typical usage.
Last edited
October 4, 2026 at 5:30 AM by Venkatraman Chandrasekaran
Similar to GPT-6.1 Sol
