OpenAI Launches GPT-6.1 Terra to Improve AI Agents and Automate Everyday Work
OpenAI has launched GPT-6.1 Terra, a new artificial intelligence model designed to help AI agents handle everyday professional tasks with greater speed, accuracy and cost efficiency. Announced on October 26, 2026, the model combines visual reasoning, computer use and natural writing to help users complete activities across digital applications. The release represents another step in OpenAI's effort to make AI agents more useful for routine workplace operations.
According to OpenAI, GPT-6.1 Terra is designed for tasks such as drafting proposals, interpreting technical drawings, completing forms, reviewing visual material and entering information into business software. The model can interpret what appears on a screen and interact with supported applications, allowing it to assist with multi-step workflows rather than simply responding to text prompts.
The company says the new model improves on GPT-5.6 Terra while reducing standard API prices. The launch is aimed at developers, businesses and professionals looking to use AI for repetitive tasks, document handling and computer-based work.
What Is GPT-6.1 Terra?
GPT-6.1 Terra is part of OpenAI's GPT-6 model family. It is designed to balance performance, speed and operating costs, particularly for tasks that require an AI system to understand visual information and take actions through a computer interface.
Traditional AI assistants primarily generate answers, summaries or other content in response to user instructions. AI agents can go further by carrying out a sequence of actions, such as reading a document, extracting relevant details, entering information into a form and preparing a report.
GPT-6.1 Terra is designed to support these workflows by combining visual understanding with computer-use capabilities. For example, a business could use the model to read incoming invoices and enter the relevant information into accounting software. A design team could use it to inspect visual assets, while an engineering team could ask it to interpret technical drawings.
The model can also help users draft and revise text, prepare business proposals and work with information displayed in applications. These capabilities are intended to reduce the manual effort involved in repetitive digital tasks.
However, the quality of the result still depends on the instructions provided, the information available to the model and the reliability of the connected applications. Human review remains important for sensitive, financial or consequential tasks.
Improved Visual Reasoning and Computer Use
One of the main features of GPT-6.1 Terra is its ability to reason about visual information. This allows the model to interpret screenshots, application interfaces, product images and technical drawings as part of a task.
Computer use allows the model to interact with supported digital environments by performing actions such as clicking, typing and navigating application screens. Combined with visual reasoning, this capability can help AI agents move through multi-step processes that would otherwise require a person to complete each step manually.
OpenAI evaluated the model using several benchmarks. On Agents' Last Exam, which tests professional tasks involving multiple steps and different types of information, the company reported that GPT-6.1 Terra scored 15 percentage points higher than GPT-5.6 Terra at medium reasoning effort, at approximately one-third of the cost.
OpenAI also reported improvements on MMMU-Pro, a benchmark that evaluates visual reasoning. On OSWorld 2.1, which tests how effectively AI agents use desktop applications to complete tasks, GPT-6.1 Terra surpassed the previous model's best reported score at medium reasoning effort.
These results suggest progress in the areas the model is designed to address. However, the figures are based on benchmark evaluations reported by OpenAI and should not be interpreted as a guarantee of the same performance across every real-world business application.
Better Factual Accuracy and Coding Performance
OpenAI has also highlighted improvements in factual accuracy. According to the company's internal evaluation using prompts designed to expose factual errors, GPT-6.1 Terra's error rate at low reasoning effort fell from 53.6 per cent for GPT-5.6 Terra to 16.7 per cent.
At maximum reasoning effort, OpenAI reported an error rate of 4.3 per cent. These figures relate to the company's specific internal test and do not mean that the model is free from mistakes. Users should continue checking important information, particularly when decisions depend on precise factual details.
The company also reported improvements in software engineering performance. On Terminal-Bench 4.0, a benchmark that evaluates how well coding agents solve command-line tasks, GPT-6.1 Terra solved 50.3 per cent of tasks at maximum reasoning effort. OpenAI compared this with a previously published score of 21.5 per cent for GPT-5.6 Terra.
OpenAI says GPT-6.1 Terra can also complete certain tasks more quickly than its more expensive models. For highly complex coding assignments, the company continues to recommend models such as GPT-6.1 Sol and GPT-6 Astra.
Lower Pricing for Developers and Businesses
Cost efficiency is another major part of the launch. OpenAI says GPT-6.1 Terra's standard API prices are 50 per cent lower than those of GPT-5.6 Terra and GPT-6.1 Sol.
The published standard API rates are one US dollar per million input tokens, five cents per million cached input tokens and five US dollars per million output tokens. Token usage measures the text and other supported content processed or generated by the model.
Lower prices can make AI automation more practical for organisations that process large volumes of information. Businesses could explore applications involving invoice processing, document review, customer support workflows, data entry and administrative tasks.
However, the total cost of an AI workflow may include more than the model's token charges. Tool calls, application integrations, infrastructure, monitoring and human verification can also contribute to the final cost. Businesses should evaluate overall task costs and accuracy rather than relying on API prices alone.
Availability of GPT-6.1 Terra
OpenAI announced that GPT-6.1 Terra is available to eligible Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex. Developers can access the model through the OpenAI API using the model identifier gpt-6.1-terra.
Access through organisational accounts may depend on workspace settings and administrator controls. Developers also need to configure their applications to use the relevant API and supported tools.
The model supports computer-use functionality through the appropriate interfaces, but practical deployment requires suitable permissions and setup. Businesses should ensure that agents have access only to the systems and information needed for their assigned tasks.
OpenAI has also described safeguards intended to keep the model within the scope of a user's instructions and improve its resistance to malicious instructions embedded in content. Such protections are important because AI agents that interact with applications may encounter untrusted documents, websites or messages. Security testing and access controls remain necessary when deploying these systems.
What the Launch Means for Workplace Automation
GPT-6.1 Terra reflects the growing focus on AI systems that can perform work rather than simply generate responses. By combining visual reasoning, computer use and writing capabilities, OpenAI is targeting tasks that occupy a significant part of many working days.
Potential applications include preparing documents, processing business records, navigating software interfaces and supporting teams that handle repetitive administrative work. These capabilities could help employees spend less time on routine steps and more time on work requiring judgement, communication and specialist knowledge.
The results will depend on how organisations integrate the model into their existing systems. Companies must consider accuracy, privacy, security, operating costs and the consequences of mistakes before allowing an agent to carry out important actions without close supervision.
GPT-6.1 Terra is therefore best understood as a tool for expanding the range of tasks AI agents can support, rather than a guarantee that all workplace processes can be automated successfully.
With its focus on computer use, visual reasoning, improved benchmark performance and lower API pricing, GPT-6.1 Terra strengthens OpenAI's offering for everyday professional workflows. Its practical impact will become clearer as developers and businesses test the model in real applications and measure whether it delivers reliable results at a lower overall cost.
