OpenAI has announced the release of its new AI model, GPT-6.1 Terra, which combines visual perception, computer interface work, and improved reasoning capabilities. The new model is aimed at automating everyday routine tasks, from document processing and drawing interpretation to video editing.
OpenAI has announced the release of its new AI model, GPT-6.1 Terra, which combines visual perception, computer interface work, and improved reasoning capabilities. The new model is aimed at automating everyday routine tasks, from document processing and drawing interpretation to video editing.
The developers' announcement states that the cost of using GPT-6.1 Terra via API is 50% lower compared to the previous versions GPT-5.6 Terra and GPT-6.1 Sol. This makes mass automation of business processes, such as reading invoices and entering them into accounting systems, much more profitable for companies and developers.
The model shows a significant increase in performance in the Agents' Last Exam test, which assesses the ability of AI to use computer programs to perform complex tasks. At an average level of reasoning, its results are 15 percentage points higher than GPT-5.6 Terra, and the cost of performing one task has decreased by about a third. The developers also note that GPT-6.1 Terra works more than three times faster than GPT-6 Astra.
A significant improvement is the reduction in hallucinations and actual errors. In OpenAI’s internal tests of complex queries, the error rate at low reasoning levels dropped from 53.6% in the previous model to 16.7% in GPT-6.1 Terra, and at maximum settings it dropped to 4.3%. In the Terminal-Bench 4.0 programming assessment, the new model solves 50.3% of command-line tasks, which is twice as high as GPT-5.6 Terra.
The model is now available to Plus, Pro, Enterprise, and Edu users on ChatGPT Work and Codex, as well as via the API. The cost for developers is $1 per 1 million inbound tokens, $0.05 per 1 million cached tokens, and $5 per 1 million outbound tokens.
OpenAI previously expanded the GPT-6 family of models with two new modifications, GPT-6 Sol and GPT-6 Luna . They were designed to provide high speed and cost savings when performing complex programming, automation, and data analysis tasks.
