OpenAI has introduced GPT-6.1 Terra, a new AI model designed to handle everyday professional tasks more efficiently, with improvements in visual reasoning, computer use, coding, and factual accuracy.
According to OpenAI, GPT-6.1 Terra is a significant upgrade over GPT-5.6 Terra, particularly for tasks that involve interacting with desktop applications, understanding visual information, and completing multistep workflows. These capabilities can help users with activities such as processing invoices, filling out forms, reviewing video footage, and turning technical drawings into 3D models.
One of the biggest changes is the model’s pricing. OpenAI says standard API pricing for GPT-6.1 Terra is 50% lower than that of GPT-5.6 Terra and GPT-6.1 Sol. The lower cost could make it more practical for developers to use AI agents for repetitive business operations, including extracting information from documents and entering data into accounting systems.
GPT-6.1 Terra also delivers stronger results on several computer-use and visual reasoning benchmarks. On Agents’ Last Exam, which measures performance on professional tasks, the model scores 15 percentage points higher than GPT-5.6 Terra at medium reasoning effort while costing about one-third as much. OpenAI also reports that Terra performs on par with GPT-6 Astra on this benchmark at roughly one-tenth of its price per token.
The model also performs well on MMMU-Pro, a benchmark designed to evaluate visual reasoning. OpenAI says GPT-6.1 Terra outperforms its predecessor at every reasoning setting and approaches the performance of more capable models, including GPT-6 Astra and GPT-6.1 Sol, at a lower cost.
Desktop automation is another area where the new model has improved. On OSWorld 2.1, which evaluates how effectively AI agents use desktop applications to complete multistep tasks, GPT-6.1 Terra surpasses the best score previously achieved by GPT-5.6 Terra at medium reasoning effort, while costing substantially less per task.
OpenAI has also focused on reducing factual errors. On its internal benchmark of difficult prompts designed to expose model weaknesses, the reported error rate at low reasoning effort dropped from 53.6% with GPT-5.6 Terra to 16.7% with GPT-6.1 Terra. At maximum reasoning effort, the error rate fell to 4.3%, an improvement of 18 percentage points over the previous model.
Coding performance has improved as well. On Terminal-Bench 4.0, which tests how well AI agents solve command-line software engineering tasks, GPT-6.1 Terra achieved a score of 50.3% at maximum reasoning effort. That is more than twice the 21.5% score reported for GPT-5.6 Terra. OpenAI says GPT-6.1 Sol and GPT-6 Astra remain better choices for the most complex coding tasks.
The model also showed strong performance on Terminal-Bench-Science 0.1, which evaluates the use of code for scientific research tasks. OpenAI reports that GPT-6.1 Terra surpasses GPT-6.1 Sol at maximum reasoning effort while costing less, and comes within five percentage points of Claude Opus 5.5 at less than one-fifth of the estimated cost.
Alongside these performance improvements, OpenAI says GPT-6.1 Terra retains the safeguards developed for GPT-6 Astra. The company reports that the model meets the Critical threshold for cybersecurity under its Preparedness Framework. Its evaluations also indicate that Terra is more likely than GPT-5.6 Terra to stay within the scope of a task, communicate limitations honestly, and resist prompt-injection attempts designed to redirect an AI agent.
GPT-6.1 Terra is available to Plus, Pro, Enterprise, and Edu users through ChatGPT Work and Codex. Developers can also access the model through the OpenAI API using the model identifier gpt-6.1-terra.
If this article helped you, please consider supporting our work. Every small contribution keeps Abijita.com independent and running.
The standard API price is $1 per million input tokens, $0.05 per million cached input tokens, and $5 per million output tokens. With its combination of lower costs, faster task completion, and improved reasoning, GPT-6.1 Terra is aimed at making AI-powered automation more accessible for everyday business and development work.
