GPT-6.1 Sol targets coding, computer use and professional workloads, while dots can run continuously on their own cloud computers and interact with external applications.
GPT-6.1 Sol specifications
The API model is available as gpt-6.1-sol.
| Specification | GPT-6.1 Sol |
| Context window | 1.05M tokens |
| Maximum output | 128K tokens |
| Knowledge cutoff | April 30, 2026 |
| Input | $2 / 1M tokens |
| Cached input | $0.10 / 1M tokens |
| Output | $10 / 1M tokens |
| Reasoning levels | Low, Medium, High, XHigh, Max |
OpenAI says GPT-6.1 Sol approaches GPT-6 Astra performance in coding, computer-use and professional-work benchmarks while costing substantially less.
Compared with GPT-6 Sol, standard API pricing remains unchanged. Cached input, however, drops from $0.20 to $0.10 per million tokens.
GPT-6.1 Sol vs GPT-6 Sol
| GPT-6 Sol | GPT-6.1 Sol | |
| Context | 1.05M | 1.05M |
| Max output | 128K | 128K |
| Input / 1M | $2 | $2 |
| Cached input / 1M | $0.20 | $0.10 |
| Output / 1M | $10 | $10 |
| Knowledge cutoff | Apr. 20, 2026 | Apr. 30, 2026 |
OpenAI reports several benchmark improvements:
- DeepSWE v1.1: up to +6.4 percentage points over GPT-6 Sol;
- AutomationBench: +4.8 points at medium reasoning;
- OSWorld 2.0: +7 points at maximum reasoning.
On OSWorld 2.0, GPT-6.1 Sol comes within 2.1 points of GPT-6 Astra while costing less than half as much per task as GPT-6 Sol at the tested settings.
OpenAI's factuality evaluation also showed fewer responses containing at least one factual error: 7.7% for GPT-6.1 Sol versus 11.4% for GPT-6 Sol at low reasoning effort.
GPT-6.1 Sol … nearly matches GPT-6 Astra’s intelligence on agentic coding, computer use, and professional work. — OpenAI
Dots: persistent GPT-6 Astra agents
OpenAI also introduced dots. Each dot is a persistent agent powered by GPT-6 Astra and equipped with its own cloud computer and browser.
A dot can:
- connect to more than 4,000 apps through ChatGPT plugins;
- work on multiple projects;
- retain project context;
- use its cloud browser and computer;
- continue tasks in the background;
- request approval for restricted actions.
Users can interact with dots through ChatGPT, Slack, Microsoft Teams and voice. SMS/text support is planned but was not available at launch.
How dots differ from ChatGPT
A normal model interaction is request-based:
Prompt → Model → Tools → Response
A dot can maintain a persistent execution loop:
Goal → Plan → Tools → Execute → Check → Continue
For software development, OpenAI says a dot can analyze bug reports, inspect code, implement fixes, run tests and prepare pull requests for review.
Each dot runs in an isolated cloud environment. Access to external applications is permission-based, and Custom Rules can allow, block or require approval for specific actions.
Background proactive research has additional restrictions: connected-app tools are read-only when the user is not actively working with the agent.
Availability
GPT-6.1 Sol is available through the OpenAI API as gpt-6.1-sol, as well as ChatGPT Work and Codex for supported paid plans.
Dots are initially rolling out to eligible Pro and Business Premium users, with Enterprise availability through a workspace-controlled beta.
The distinction for software engineers is clear - to build agentic applications at a lower price, developers use GPT-6.1 Sol. In contrast OpenAI provides the persistent agent infrastructure as a completed product called dots. There is a specific purpose for each tool. By using the first model, the individuals who write code lower expenses - but the second option is ready for immediate use. With those tools OpenAI offers different ways to manage automated systems. It is possible to select the model that fits the budget or the finished infrastructure. The two options are distinct for the who create software.
