GPT-6.1 Terra: OpenAI's Cheap Computer-Use Workhorse
GPT-6.1 Terra is OpenAI's new fast, cheap model for computer use and coding, priced at $1 and $5 per million tokens and live in ChatGPT Work, Codex and the API.
- OpenAI says standard API prices are 50% lower than both GPT-5.6 Terra and GPT-6.1 Sol.
- On its internal error-inducing prompts, the error rate at low reasoning effort fell from 53.6% with GPT-5.6 Terra to 16.7%.
- AWS announced GPT-6.1 Terra is generally available on Amazon Bedrock too, so it is not locked to OpenAI's own platform.
in this block
GPT-6.1 Terra just dropped, and it is OpenAI's cheap, fast workhorse for computer use. OpenAI announced the model on Monday, October 26, calling it "a significant step up from GPT-5.6 Terra" for visual reasoning, form-filling, writing and coding. It is live in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users, and in the API as gpt-6.1-terra at $1 per million input tokens and $5 per million output tokens.
What actually happened
OpenAI's post pitches GPT-6.1 Terra for "the communication and computer-based tasks that fill the workday," like drafting proposals, reading technical drawings, filling out forms and editing videos. The headline is cost: OpenAI says Terra gives "a new balance of speed, capability, and cost," and that the same ChatGPT usage allowance now covers more tasks than with GPT-5.6 Terra.
Computer use is the big flex. On Agents' Last Exam, a benchmark of professional tasks, OpenAI says Terra at medium reasoning scores 15 percentage points higher than GPT-5.6 Terra at about one-third the cost. It says Terra performs on par with GPT-6 Astra at one-tenth the price per token. At max effort, OpenAI claims Terra is over three times as token-efficient as Claude Opus 5.5 and one-fourth its price per token, for about 9x lower cost per task.
Speed is part of it too. OpenAI says Terra finishes tasks over three times faster than both GPT-6 Astra and GPT-6.1 Sol.
Benchmarks worth a closer look
Coding got a real bump, according to OpenAI. On Terminal-Bench 4.0 at maximum effort, GPT-6.1 Terra solves 50.3% of tasks, more than double GPT-5.6 Terra's published 21.5%. On Terminal-Bench-Science 0.1, OpenAI says Terra beats GPT-6.1 Sol at lower cost and comes within five points of Opus 5.5.
OpenAI is also honest about limits. It says "for the most complex coding tasks, GPT-6.1 Sol and GPT-6 Astra remain the better choice." So Terra is not the new frontier king. It is the model you run a thousand times a day without wincing at the bill.
Factuality is the other big claim. Beyond the low-effort drop to 16.7%, OpenAI says the max-effort error rate fell 18 points to 4.3%, nearly matching GPT-6 Astra and GPT-6.1 Sol. These are OpenAI's internal tests, so wait for independent evals before treating them as settled.
Frontier models get the headlines. Cheap models that are good enough get the traffic.
Safety and the cyber flag
One line deserves attention: OpenAI says GPT-6.1 Terra meets the "Critical" threshold in cybersecurity under its Preparedness Framework and keeps the protections built for GPT-6 Astra. In plain terms, OpenAI rates it as capable enough in cyber tasks to need strong safeguards.
OpenAI also says Terra is more likely to stay in scope and disclose blockers than GPT-5.6 Terra, and that it showed strong resistance in prompt-injection tests. That matters a lot for agents that click around real apps. Details are in the system card.
Where this fits
We tracked the run-up in our GPT Terra release odds piece, and the pricier sibling in GPT-6.1 Sol. With Terra at $1 and $5 per million tokens, OpenAI is pushing hard into high-volume agent work where rivals like Claude Opus 5.5 are strong but expensive. This release is the price-performance shot in that fight.
The Bedrock listing matters more than it looks. Big companies that already run on AWS can now try the model inside their existing cloud contracts, security reviews and billing. That removes a lot of friction for enterprise pilots, and it shows how much OpenAI wants this model used at volume, not just tested in a chat window.
What to do as a reader
If you build agents, rerun your evals on gpt-6.1-terra today. Use the same prompts, tools and time limits you use for your current model so the comparison is fair. Check cost per finished task, not just cost per token, because OpenAI's pitch is that Terra uses fewer tokens too. Read the OpenAI announcement and the AWS Bedrock note for availability details.
If you use ChatGPT Work or Codex, try the new model for routine tasks and keep Sol or Astra for hard coding. There is no token here. Any "Terra AI coin" on a DEX has nothing to do with OpenAI.
Not financial advice. DYOR, ser.
