OpenAI launches GPT-6.1 Terra, a cheaper AI model
OpenAI introduces GPT-6.1 Terra, a model that combines visual understanding, application use and programming at a lower cost. It is available on several ChatGPT plans and to developers through the API, with prices 50% lower than previous models and notable improvements in accuracy and speed.
OpenAI has launched GPT-6.1 Terra, a model designed to handle work tasks involving images, applications and text at a lower cost than its more powerful models. It can interpret technical drawings, complete forms, edit videos and draft proposals in more natural-sounding prose.
The main development is not just that Terra can do more. It is that it tries to do those tasks faster and more cheaply, which matters when a company needs to automate thousands of tasks a day.
More capacity to use your computer
GPT-6.1 Terra improves how it uses desktop applications and digital tools. In practice, it can read an incoming invoice and enter its data into accounting software, turn an architectural drawing into a 3D model or review a film frame by frame to create a log with timecodes.
OpenAI says that, in its Agents’ Last Exam evaluation, Terra scores 15 percentage points higher than GPT-5.6 Terra with a medium reasoning level, at approximately one-third of the cost. The company also says it matches GPT-6 Astra’s performance in that test at one-tenth of the price per token.
A token is a unit of text that the model processes to read instructions and generate responses. Its price matters especially when an application performs thousands or millions of operations.
Terra also completes these tasks faster: OpenAI says it finishes jobs more than three times faster than GPT-6 Astra and GPT-6.1 Sol. In the OSWorld 2.1 test, which measures how an AI handles desktop applications in multi-step processes, it beats GPT-5.6 Terra’s best published score at a much lower cost per task.
Fewer errors and better code
OpenAI also highlights a reduction in erroneous responses in its internal evaluation of instructions designed to trigger failures. With a low reasoning level, the error rate fell from 53.6% to 16.7% compared with GPT-5.6 Terra. At the highest level, it stood at 4.3%, according to the company’s data.
In programming, GPT-6.1 Terra scores 50.3% on Terminal-Bench 4.0 with the highest reasoning level, compared with the 21.5% published for GPT-5.6 Terra. This test measures whether an agent can solve problems from the command line, such as fixing a page whose layout changes while it loads.
Even so, OpenAI maintains a clear distinction: for more complex programming work, GPT-6.1 Sol and GPT-6 Astra remain more suitable options.
Price and availability
GPT-6.1 Terra is available today to users on the Plus, Pro, Enterprise and Edu plans in ChatGPT Work and Codex. Developers can use it through the API under the name gpt-6.1-terra.
The standard API pricing is:
- $1 per one million input tokens.
- $0.05 per one million cached input tokens.
- $5 per one million output tokens.
According to OpenAI, these rates are 50% lower than those for GPT-5.6 Terra and GPT-6.1 Sol. The goal is to make automating repetitive tasks at scale viable for more companies, not just those that need the most powerful model.
GPT-6.1 Terra also includes the GPT-6 series’ safety improvements. OpenAI says it better respects the scope of a task, recognizes its limitations more clearly and resists malicious instructions that try to divert it from its objective. The company also says it meets the critical cybersecurity threshold in its preparedness framework.
What remains to be seen is how it performs outside OpenAI’s tests. Terra’s important advance is its combination of vision, computer use, programming and a lower price in one model. For you, that could mean assistants capable of completing more steps on their own, as long as companies keep controls in place to review their decisions and correct their mistakes.