OpenAI has introduced GPT-6.1 Terra, a model designed to handle everyday professional work involving software, documents, images and browser-based applications. The upgrade builds on GPT-5.6 Terra with stronger visual understanding, more precise computer interactions and improvements in writing and coding performance.
The model is aimed at businesses and developers that want AI systems to do more than answer questions. By combining improved task execution with lower API prices, OpenAI is positioning GPT-6.1 Terra as a practical option for automating high-volume workflows, from processing invoices to navigating applications and updating business records. It is available through ChatGPT Work, Codex and the OpenAI API for eligible users and developers.
Key Takeaways
- New model: GPT-6.1 Terra launched on October 26, 2026.
- Lower API prices: Standard API pricing is 50% lower than GPT-5.6 Terra and GPT-6.1 Sol.
- Computer use: The model can interpret screens and interact with software, including filling out forms and working with spreadsheets.
- Improved coding: OpenAI reports a Terminal-Bench 4.0 score of 50.3% at maximum reasoning effort, compared with 21.5% for GPT-5.6 Terra.
- Faster execution: OpenAI says Terra completes computer-based tasks more than three times faster than GPT-6 Astra and GPT-6.1 Sol in its reported comparisons.
- Availability: Developers can access the model through the API, while eligible paid users can use it in ChatGPT Work and Codex.
What Is GPT-6.1 Terra?
GPT-6.1 Terra is a model in OpenAI’s GPT-6 generation, built to balance performance, speed and cost for routine professional work.
Its main focus is computer use: the ability to understand what is displayed on a screen and interact with software through actions such as clicking, typing and navigating interfaces. This makes it useful for tasks that traditionally require a person to move between applications and manually enter information.
For example, an AI agent could read invoice details, open an accounting application, enter the relevant fields and check whether the information was recorded correctly. Another workflow might involve reviewing a spreadsheet, identifying inconsistencies and preparing a summary.
Terra also supports visual reasoning and natural writing. OpenAI highlights applications such as interpreting technical drawings, drafting proposals and reviewing video footage.
The model is not intended to replace every model in OpenAI’s lineup. Instead, it targets work where a capable model needs to complete many practical tasks efficiently, without requiring the highest-cost option for every request.
What Does 50% Lower Pricing Mean?
OpenAI says GPT-6.1 Terra’s standard API prices are 50% lower than those of GPT-5.6 Terra and GPT-6.1 Sol.
The published rates are:
| API usage | GPT-6.1 Terra price |
|---|---|
| Input tokens | $1 per million |
| Cached input tokens | $0.05 per million |
| Output tokens | $5 per million |
Source: OpenAI’s official model documentation.
Input tokens represent the information sent to the model, while output tokens represent the content it generates. Cached input receives a lower rate when eligible input can be reused.
The lower price could be significant for applications that process thousands of documents or perform repeated computer interactions. A company handling large volumes of invoices, for example, may need an AI model to extract information, make decisions and populate software fields throughout the day.
However, token prices alone do not determine the total cost of automation. Computer-use tool calls, retries, execution time and human review can also affect expenses. Developers should measure the full cost of completing a task rather than comparing API prices in isolation.
Improved Computer Use and Visual Reasoning
Computer use is one of the central features of GPT-6.1 Terra. The model is designed to interpret visual information and interact with applications rather than relying exclusively on text-based inputs.
This capability could support several workflows:
- Browser automation: Navigate websites and complete forms.
- Spreadsheet work: Read information, update fields and assist with data analysis.
- Document processing: Extract information from invoices, proposals and business documents.
- Technical interpretation: Examine drawings and other visual materials.
- Video review: Analyze footage and prepare time-coded notes for further work.
These capabilities are particularly useful when an organization relies on software that does not provide a convenient API for every task. An agent that can interact with a graphical interface may automate some work that would otherwise require a person to operate the application manually.
Computer-use systems still require safeguards. An agent can click the wrong control, misread a screen or make an incorrect assumption about what a button does. Tasks involving payments, sensitive records or irreversible changes should use appropriate permissions, verification and human confirmation.
GPT-6.1 Terra Shows Improvements in Coding Benchmarks
Although Terra emphasizes everyday professional work, OpenAI also reports a substantial improvement in coding performance.
On Terminal-Bench 4.0, a benchmark that evaluates agents completing command-line tasks, GPT-6.1 Terra achieved a score of 50.3% at maximum reasoning effort. GPT-5.6 Terra scored 21.5% under the reported comparison.
That is an improvement of 28.8 percentage points. It suggests that the updated model is better equipped to handle tasks involving command-line tools and software workflows.
OpenAI also reports improvements on Agents’ Last Exam, a benchmark of professional computer-based tasks. At medium reasoning effort, GPT-6.1 Terra scored 15 percentage points higher than GPT-5.6 Terra while costing about one-third as much per task.
These results are encouraging, but they should be interpreted carefully. Benchmark scores measure performance on specific test sets, not every possible programming or office task. Real-world results depend on the software environment, instructions, available tools and complexity of the work.
For demanding coding projects, OpenAI continues to recommend GPT-6.1 Sol or GPT-6 Astra when deeper reasoning is required.
A Focus on Accuracy and Safer Task Execution
OpenAI also reports a reduction in factual errors on its internal evaluation using prompts designed to induce mistakes.
At low reasoning effort, the reported error rate fell from 53.6% with GPT-5.6 Terra to 16.7% with GPT-6.1 Terra. At maximum reasoning effort, OpenAI reports an error rate of 4.3% for the new model.
The company also highlights improved alignment, including a greater tendency to remain within the requested task and disclose limitations or obstacles. Its announcement describes stronger resistance to prompt-injection attempts, in which malicious instructions try to redirect an AI agent away from the user’s original goal.
These results are relevant because computer-use agents can affect real applications and data. A model that follows instructions more reliably may reduce mistakes, but it cannot eliminate them. Independent testing and external safeguards remain important, particularly when agents can access sensitive information or perform consequential actions.
How Businesses Could Use GPT-6.1 Terra
The combination of computer use and lower pricing creates opportunities for businesses that want to automate repetitive workflows.
| Business area | Potential application |
|---|---|
| Finance and accounting | Extract invoice details and enter them into accounting systems |
| Customer support | Review requests, update records and prepare responses |
| Sales operations | Update customer relationship management systems and organize lead information |
| Software development | Execute command-line tasks, run tests and help debug code |
| Research and administration | Review documents, compare information and prepare summaries |
| Media and design | Analyze visual materials and assist with video-review workflows |
The most promising applications are likely to be repetitive tasks with clearly defined steps and measurable outcomes. Organizations can begin by automating low-risk workflows, checking results against known answers and expanding access as reliability improves.
For companies already using AI agents, the lower API price could also make it economical to apply automation to tasks that previously cost too much to process at scale.
The Bigger Picture
GPT-6.1 Terra reflects a broader shift in AI development from conversational systems toward agents that can operate software and complete multi-step work. As models become better at interpreting screens and using tools, the boundary between an AI assistant and a digital worker becomes less distinct.
Lower pricing could accelerate this shift by reducing the cost of running frequent, routine tasks. However, the strongest business cases will depend on more than model performance: organizations must also account for integration costs, error handling, privacy, security and the need for human oversight.
For OpenAI, Terra strengthens the middle of its model lineup by combining practical computer-use capabilities with more accessible API pricing. Its adoption will depend on whether developers can translate benchmark improvements into measurable savings and reliable results in everyday operations.
FAQs
1. What is GPT-6.1 Terra?
GPT-6.1 Terra is an OpenAI model focused on visual reasoning, computer use, coding and natural writing. It is designed to handle everyday professional tasks efficiently.
2. Is GPT-6.1 Terra 50% cheaper?
OpenAI says its standard API prices are 50% lower than GPT-5.6 Terra and GPT-6.1 Sol. The published rates are $1 per million input tokens and $5 per million output tokens, with eligible cached input priced at $0.05 per million tokens.
3. Can GPT-6.1 Terra use a computer?
Yes. The model supports computer-use workflows, allowing agents to interpret screens and interact with applications. Developers must still configure the necessary tools, permissions and safeguards.
4. Is GPT-6.1 Terra good for coding?
OpenAI reports a Terminal-Bench 4.0 score of 50.3% at maximum reasoning effort, compared with 21.5% for GPT-5.6 Terra. For the most demanding coding and reasoning tasks, OpenAI still points users toward GPT-6.1 Sol or GPT-6 Astra.
5. Where can developers access GPT-6.1 Terra?
Developers can use the model through the OpenAI API under the model ID gpt-6.1-terra. Eligible paid users can also access it through ChatGPT Work and Codex.
Looking Ahead
The next test for GPT-6.1 Terra will be its performance in production workflows. Businesses will need to determine whether the model’s improved computer use and lower token prices translate into fewer manual steps, shorter completion times and lower overall operating costs.
As AI agents take on more software-based work, competition will increasingly center on the combination of capability, cost and reliability. GPT-6.1 Terra is OpenAI’s latest effort to make that combination more practical for everyday automation, while leaving the most complex reasoning tasks to its higher-capability models.
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
