OpenAI Launches GPT-6.1 Terra: A Faster, Cheaper AI Model for Everyday Work

OpenAI Launches GPT-6.1 Terra: A Faster, Cheaper AI Model for Everyday Work

OpenAI is making another move in the AI race, but this time it's not just about building a smarter chatbot. It is about making AI more useful for everyday work without making businesses spend a fortune.
 

OpenAI introduced GPT-6.1 Terra, a new model designed to combine visual understanding, computer use, improved accuracy, and natural writing at a lower cost. The model is built for tasks that take up much of the working day, including filling out forms, interpreting technical drawings, drafting proposals, reviewing documents, and editing videos.

In simple terms, OpenAI wants AI to become better at understanding what is happening on a computer screen and taking useful actions based on that information. And for businesses looking to automate repetitive work, that could be a significant development. 

What Makes GPT-6.1 Terra Different?

One of Terra's biggest strengths is its ability to understand visual information and interact with computer applications.

For example, rather than simply explaining what is shown in a document, it can help interpret the information and take the next steps within supported applications. It can also work with technical drawings, navigate software interfaces, and complete tasks that require multiple actions.

This is particularly useful for AI agents. These are AI systems designed to carry out tasks, not just respond to questions.

A business could use this kind of technology to process incoming invoices, extract important information, and enter the details into accounting software. Another company could use it to review documents, organize information or automate repetitive administrative processes.

GPT-6.1 Terra also brings improvements in writing, making its responses clearer and more natural. The bigger picture is that AI is gradually moving from being a tool people consult to becoming a tool that can help them get actual work done.

OpenAI Says It Makes Fewer Mistakes

Another important improvement is accuracy.

One of the biggest challenges with AI is that models can sometimes provide incorrect information confidently. OpenAI says it has invested in reducing these factual errors in GPT-6.1 Terra. According to the company's internal evaluations, the error rate on difficult prompts dropped from 53.6% with GPT-5.6 Terra to 16.7% with GPT-6.1 Terra at low reasoning effort. At maximum reasoning effort, the reported error rate fell to 4.3%.

Coding performance also improved. On Terminal-Bench 4.0, a test that evaluates how well AI systems solve software engineering tasks, GPT-6.1 Terra achieved 50.3% at maximum reasoning effort, compared with 21.5% for GPT-5.6 Terra's published score.

These results are based on OpenAI's reported evaluations, so they should be understood as company-reported benchmarks rather than guarantees of performance in every real-world situation.

OpenAI still recommends its more capable models, including GPT-6.1 Sol and GPT-6 Astra, for the most demanding coding and reasoning tasks.

The Big Advantage? It Costs Less

Here is where GPT-6.1 Terra becomes particularly interesting.

OpenAI says its standard API prices are 50% lower than those of GPT-5.6 Terra and GPT-6.1 Sol.

The pricing is:

Input: $1 per million tokens.

Cached input: $0.05 per million tokens.

Output: $5 per million tokens.

For developers and businesses processing large volumes of information, lower AI costs could make automation more practical. Think about a company handling thousands of invoices, customer requests, or business documents every month. Using an expensive model for every routine task can quickly increase operating costs.

A faster, more affordable model that can still handle these tasks effectively gives businesses another option. It also means developers can explore building AI-powered products and workflows without necessarily relying on the most expensive models available.

Where Can You Access GPT-6.1 Terra?

GPT-6.1 Terra is rolling out to Plus, Pro, Business, Enterprise, and Edu users through ChatGPT Work and Codex, subject to account and workspace availability. It is important to note that Terra is available in these work-focused experiences, not the regular ChatGPT chat interface.

Developers can also access it through the OpenAI API using the model name gpt-6.1-terra.

Additionally, Amazon Web Services announced its general availability on Amazon Bedrock, giving eligible developers and businesses another way to use the model.

OpenAI Is Making the Economics of AI a Priority

The release of GPT-6.1 Terra highlights an important shift in the AI industry.

Companies are no longer competing only on how intelligent their models can become. They are also competing on speed, reliability, affordability, and how much real work their AI systems can complete.

For businesses, the most useful AI model may not always be the biggest or most expensive one. It may be the model that completes everyday tasks accurately, quickly, and at a cost that makes financial sense.

GPT-6.1 Terra is OpenAI's latest attempt to deliver that balance.

As AI adoption continues to grow, making powerful capabilities more affordable could be just as important as making the models smarter.