OpenAI made task automation cheaper with new GPT-6.1 Terra

American company OpenAI presented GPT-6.1 Terra - a new artificial intelligence model focused on automating everyday work processes. It is capable of analyzing images, interacting with computer programs, working with documents, and performing programming tasks. At the same time, its use via API will cost twice cheaper than the previous Terra model.

The company announced this on October 26.

The main advantage of GPT-6.1 Terra is the ability to perform multi-step tasks directly in computer applications. The model can analyze screen content, work with forms, interpret technical drawings, prepare documents, and assist with video editing.

For business, this opens up opportunities for automating operations that previously required a significant amount of manual work. For example, an AI agent can read data from supplier invoices and transfer them to an accounting system, as well as perform sequences of actions in corporate applications.

How much cheaper is AI use

Standard rates for GPT-6.1 Terra via API are 50% lower than for GPT-5.6 Terra and GPT-6.1 Sol. The cost is $1 per million input tokens and $5 per million output tokens. Processing a million cached input tokens costs $0.05.

Tokens are units of text information by which the cost of language models is calculated. The rate reduction is particularly important for companies that process large volumes of documents daily or use AI for automatic execution of repetitive tasks.

According to OpenAI's internal tests, in the Agents' Last Exam the new model surpassed GPT-5.6 Terra by 15 percentage points at moderate reasoning effort. At the same time, the estimated cost of task completion remained about three times lower.

The company also states that in this test GPT-6.1 Terra completed tasks more than three times faster than more powerful GPT-6 Astra and GPT-6.1 Sol.

Fewer errors and better programming results

Developers specifically highlight improved factual accuracy. In an internal test with complex queries specially selected to check propensity for errors, the share of incorrect answers at low reasoning level decreased from 53.6% in the previous Terra to 16.7% in the new model.

Significant progress has also been achieved in programming. At maximum reasoning effort GPT-6.1 Terra completed 50.3% of tasks in Terminal-Bench 4.0 compared to 21.5% for GPT-5.6 Terra.

At the same time, OpenAI acknowledges that for the most complex programming tasks and deep research, GPT-6.1 Sol and GPT-6 Astra remain more powerful.

The claimed improvements in speed, accuracy, and cost were obtained in specific test conditions. They do not imply that the new model will have the same advantage in every real-world working scenario.

Who has access to GPT-6.1 Terra

The new model became available to users of ChatGPT Plus, Pro, Enterprise, and Edu plans in ChatGPT Work and Codex services.

Developers can connect via the OpenAI API under the name gpt-6.1-terra.

In ChatGPT Work and Codex, using Terra also allows performing more tasks within the limits set by the plan compared to more powerful models.

Thus, OpenAI is betting not only on increasing AI's intellectual capabilities, but also on reducing the cost of its practical application. For corporate clients, this may expand the list of processes whose automation becomes economically feasible.

Based on materials from: OpenAI, BleepingComputer, OpenAI (model technical documentation)