OpenAI GPT-6.1 Terra can see your screen and act across apps. Here is how
OpenAI has launched GPT-6.1 Terra, a lower-cost artificial intelligence model built to understand images and screens, reason about what it sees and carry out tasks across computer applications. The model became available on 26 October in ChatGPT Work, Codex and the OpenAI API.
Terra targets everyday professional work where speed and cost matter. OpenAI says it can read technical drawings, complete forms, edit videos, handle invoices and produce clearer writing. These are company claims based largely on internal tests, so performance can vary across software, permissions and real-world workflows.
:
What GPT-6.1 Terra can do
The model combines visual reasoning with computer use. It can inspect an application screen, identify relevant controls and perform a sequence of authorised actions. This could help with repetitive jobs such as matching product images to catalogue records or entering information from documents into business software.
Computer use still depends on supported tools, account access and user-approved permissions. Organisations should review important actions, especially when a task involves payments, confidential files or changes that are difficult to reverse.
:
How Terra compares with earlier models
OpenAI describes GPT-6.1 Terra as a major upgrade over GPT-5.6 Terra. On Agents’ Last Exam, its benchmark for professional software tasks, the new model scored 15 percentage points higher at medium reasoning effort while using about one-third of the cost per task.
On Terminal-Bench 4.0, Terra solved 50.3 per cent of tasks at maximum reasoning effort. GPT-5.6 Terra’s published score was 21.5 per cent. OpenAI still recommends GPT-6.1 Sol or GPT-6 Astra for the most complex coding work.
The company also reports fewer factual errors. On an internal set designed to trigger mistakes, the error rate at low reasoning effort fell from 53.6 per cent to 16.7 per cent. Independent testing across broader tasks will be needed to judge how consistently that improvement appears.
:
GPT-6.1 Terra: Price, access and safety
Developers can use the model under the name gpt-6.1-terra. Standard API rates are $1 per million input tokens, $0.05 for cached input and $5 per million output tokens. OpenAI says those standard rates are 50 per cent below GPT-5.6 Terra and GPT-6.1 Sol.
The API documentation lists a 1.05 million-token context window, a 128,000-token maximum output and a 1 June 2026 knowledge cutoff. Tool calling requires the Responses API.
Plus, Pro, Enterprise and Edu subscribers can access Terra through ChatGPT Work and Codex as rollout reaches their accounts. Enterprise availability may also depend on administrator controls.
OpenAI classifies the model at the Critical threshold for cybersecurity under its Preparedness Framework and says it carries protections developed for GPT-6 Astra.