GPT-6.1 Terra model halves OpenAI API prices and targets computer use

OpenAI used Monday, the day Anthropic filed for its IPO, to ship a cheaper workhorse. The GPT-6.1 Terra model is pitched as the default for everyday agent work, from filling in forms and reading invoices to editing video, at a price meant to make high-volume automation pay.

What happened

Terra is available now to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, and to developers through the API as gpt-6.1-terra, OpenAI said. The Verge described it as a model OpenAI says is very good at computer use. Developers pay $1 for each million input tokens, five cents for each million cached ones and $5 for each million output tokens, half what GPT-5.6 Terra and GPT-6.1 Sol cost.

OpenAI's benchmark claims are aggressive. On Agents' Last Exam, a test of professional computer tasks, Terra at medium reasoning effort scored 15 percentage points above GPT-5.6 Terra at about a third of the cost, and matched GPT-6 Astra at a tenth of its per-token price. At maximum effort it was more than three times as token-efficient as Anthropic's Claude Opus 5.5, at a quarter of the price per token, for about nine times lower cost per task, the company said.

Why it matters

Reliability is the bigger selling point for agents. OpenAI says its own test of prompts designed to trigger mistakes showed errors dropping to 16.7% at low effort, versus 53.6% for GPT-5.6 Terra, and to 4.3% at maximum effort. On Terminal-Bench 4.0, a command-line coding test, the model solved 50.3% of tasks against 21.5% for its predecessor, though OpenAI said Sol and Astra remain better for the hardest coding work.

The release extends a price war at the agent layer. Cheaper tokens let companies run more autonomous steps per dollar, which matters more than peak benchmark scores for back-office automation. We covered earlier steps in our reports on GPT-6 Astra in contracting and the GPT-6 interface rollout. The GPT-6.1 Terra model also lands as investors weigh how much pricing power frontier labs keep while racing to cut costs.

There is a safety note too. OpenAI said Terra meets the Critical threshold for cybersecurity under its Preparedness Framework and carries the protections built for GPT-6 Astra. The claims are vendor-run; independent tests have yet to confirm them.

OpenAI also said the model is more likely than GPT-5.6 Terra to stay within the scope of a task and to flag its own limitations or blockers, and that it showed strong resistance in prompt-injection tests, where malicious text tries to hijack an agent. For companies letting agents click through internal systems, those properties may matter as much as raw speed, which OpenAI put at more than three times faster than GPT-6 Astra and GPT-6.1 Sol on tasks.

What's next

Developers will now test whether the GPT-6.1 Terra model holds up outside OpenAI's own evaluations, especially on long, multi-step tasks. Rivals are likely to answer on price. Benchmarks and pricing are on OpenAI's announcement page.