OpenAI Releases GPT-6.1 Terra Model to Enhance AI Agent Capabilities
October 30, 2026 · 2 min read
OpenAI has released GPT-6.1 Terra, a new model designed to support AI agents in tasks that require the understanding of screens, documents, and other visual content. The model also supports basic level development and programming activities.
The model is designed to expand the use of artificial intelligence in daily business processes, including document management and interaction with work tools and applications. It features computer use capabilities, allowing the model to understand what appears on a screen and perform operations within applications using authorized tools. Examples include reading invoice information to enter it into a management system or comparing product images with catalog sheets to update them.
Benchmark results indicate superior performance compared to the previous generation. In the Agents Last Exam benchmark, which measures the operational capabilities of AI agents, GPT-6.1 Terra achieved a 15 percentage point improvement over GPT-5.6 Terra at the Medium reasoning level. In the Terminal-Bench 4.0 benchmark for command line development tasks, the model with Max reasoning solved 50 percent of tasks, compared to 21.5 percent for GPT-5.6 Terra. OpenAI continues to recommend GPT-6.1 Sol and GPT-6 Astra for more advanced coding projects.
OpenAI reports a reduction in factual errors. In a test using prompts designed to provoke wrong answers with low reasoning, the error rate dropped to 16.7 percent from the 53.6 percent recorded by GPT-5.6 Terra.
Regarding efficiency, execution times in presented tests are more than three times faster than GPT-6.1 Sol and GPT-6 Astra. The model reaches performance comparable to GPT-6 Astra while maintaining a cost per token that is approximately ten times lower.
Standard API costs for GPT-6.1 Terra are 1 dollar per million input tokens and 5 dollars per million output tokens, with 0.05 dollars per million cached input tokens. These standard rates are 50 percent lower than those of GPT-5.6 Terra and GPT-6.1 Sol. Additional costs may apply for cache writing, tools, and requests exceeding 272,000 input tokens.
GPT-6.1 Terra is available via API, Codex, and ChatGPT Work for Plus, Pro, Business, Enterprise, and Edu plans. It is not available in the standard ChatGPT Chat mode. The model features a context window of up to 1,050,000 tokens and a maximum output of 128,000 tokens.
Under its Preparedness Framework, OpenAI has assigned the model a Critical capability classification for cybersecurity. The company stated it has adopted the same safeguards used for GPT-6 Astra and GPT-6.1 Sol.