OpenAI Launches GPT-6.1 Terra, a Computer-Use Model at Half the Price of Its Predecessor

OpenAI has introduced GPT-6.1 Terra, a model designed to work on a computer: it understands what is on the screen and carries out actions across different applications. It launched on October 26, 2026, and according to OpenAI its API prices are half those of GPT-5.6 Terra.

What Terra can do

OpenAI describes it as a fast, efficient model that combines detailed visual reasoning and precise computer use with clearer, more natural writing. It is meant to fill out forms, edit videos, interpret technical drawings and prepare proposals. The company's example of high-volume work is reading incoming invoices and entering their details into accounting software.

According to OpenAI, the model is a clear step up from GPT-5.6 Terra and costs far less than the company's larger models, GPT-6.1 Sol and GPT-6 Astra. OpenAI still recommends Sol and Astra for the most complex programming tasks, so Terra is not positioned as a replacement for them in hard coding.

Pricing and results

According to OpenAI, API rates are 50% lower than for GPT-5.6 Terra and GPT-6.1 Sol: $1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens.

OpenAI says that on Agents' Last Exam at medium reasoning effort, Terra scores 15 percentage points higher than GPT-5.6 Terra at about one-third the cost. The company also claims the model matches GPT-6 Astra at one-tenth the price per token. These are the vendor's own measurements, not independent tests.

It even performs on par with GPT-6 Astra at only one-tenth its price per token. - OpenAI, blog post quoted by BleepingComputer

On OpenAI's internal benchmark at maximum effort, the error rate fell by 18 percentage points to 4.3%, which the company says nearly matches the rates for GPT-6 Astra and GPT-6.1 Sol. The sources cite no independent tests of the model.

Critical level in cybersecurity

Help Net Security reports that, according to OpenAI, Terra meets the Critical threshold for cybersecurity under the Preparedness Framework version dated April 2025. By definition, a model reaches that level if, with tools and no human involved, it can find and build working exploits for zero-day flaws of all severity levels in many hardened real-world critical systems, or if it can plan and run a new attack from start to finish against a hardened target given only a high-level goal.

Terra keeps the protections built for GPT-6 Astra, which on September 1, 2026 became the first OpenAI model designated as Critical. The company says Terra held up well in tests with malicious instructions meant to pull the model off the user's task, and that it is more likely than its predecessor to stay within scope and to disclose limitations honestly.

What it means for Polish companies

For teams automating office work such as invoice or form processing, the main point is the price drop alongside claimed quality gains. The results come mostly from OpenAI, though, so it is worth testing the model on your own documents before deploying it. The model is already available to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, and in the API as gpt-6.1-terra.