What Is GPT-6.1 Terra? OpenAI's Cheaper Computer-Use Model

OpenAI has rolled out GPT-6.1 Terra, a fast, cheaper model built for computer use, visual reasoning, coding, and everyday office work. It replaces GPT-5.6 Terra, cuts standard API prices in half, and is available now in ChatGPT Work, Codex, and the API as gpt-6.1-terra. If you automate high-volume, computer-based tasks, here's what actually changed, the benchmark claims, and when Terra beats Sol or Astra.

What GPT-6.1 Terra is

Terra is the mid-tier model in OpenAI's lineup. The company organizes its main model classes from lowest price and capability to highest as Luna, Terra, Sol, and Astra. When OpenAI shipped the GPT-6 generation of Sol and Luna last month, Terra was left on the older GPT-5.6 version — this release brings it up to GPT-6.1.

OpenAI describes Terra as a model that combines detailed visual reasoning and precise computer use with clearer, more natural writing. It's pitched at the tasks that fill a workday: drafting proposals, interpreting technical drawings, filling out forms, reading invoices into accounting software, and editing videos.

The headline positioning is cost-efficiency. Terra is meant to do more work for longer without the price tag of the larger GPT-6 Astra or GPT-6.1 Sol.

Pricing: 50% lower than GPT-5.6 Terra

The biggest change is cost. Standard API pricing for GPT-6.1 Terra is:

OpenAI says these rates are 50% lower than both GPT-5.6 Terra and GPT-6.1 Sol, which lowers the direct inference cost of running the model in production. In ChatGPT Work and Codex, the same usage allowance covers more tasks than GPT-5.6 Terra, and substantially more than GPT-6.1 Sol or GPT-6 Astra.

Benchmarks: faster and cheaper, not a flagship

OpenAI's pitch for Terra is price-per-task and speed, not topping the charts. The vendor-reported numbers:

• Agents' Last Exam (how well agents use software at work): at medium reasoning effort, OpenAI says Terra scores about 15 percentage points above GPT-5.6 Terra at roughly one-third the cost, and matches GPT-6 Astra at one-tenth the price per token.
• Speed: Terra reportedly completes tasks more than 3× faster on average than both GPT-6 Astra and GPT-6.1 Sol.
• Terminal-Bench 4.0 (coding): at maximum reasoning effort, OpenAI reports a 50.3% task-completion rate, versus a published 21.5% for GPT-5.6 Terra.
• MMMU-Pro and OSWorld 2.1: OpenAI says Terra beats GPT-5.6 Terra across reasoning settings on MMMU-Pro, and on OSWorld 2.1 medium-effort Terra exceeds GPT-5.6 Terra's best score at a fraction of the cost.

OpenAI also compares Terra to Anthropic's Claude Opus 5.5: at maximum effort, the company claims Terra is over three times as token-efficient and one-fourth the price per token, which it frames as roughly 9× lower cost per task.

A caveat worth keeping: these are all vendor-published figures from the launch, so treat the cross-model comparisons as OpenAI's own claims until third parties run Terra head-to-head. And OpenAI itself says GPT-6.1 Sol and GPT-6 Astra remain the better pick for the most complex coding work.

Fewer factual errors

OpenAI has spent this year reducing hallucinations, and Terra reflects that. On the company's internal benchmark of error-inducing prompts, the error rate at low reasoning effort fell from 53.6% with GPT-5.6 Terra to 16.7% with GPT-6.1 Terra — a drop of nearly 37 percentage points. At maximum effort, the error rate fell by 18 points to 4.3%, nearly matching GPT-6 Astra and GPT-6.1 Sol.

OpenAI also reports closer scope adherence, more honest disclosure of its own limitations and blockers than GPT-5.6 Terra, and resistance to prompt injection in its tests.

Safety classification

OpenAI classifies GPT-6.1 Terra at the "Critical" cybersecurity threshold under its Preparedness Framework and says the model retains the safeguards used for GPT-6 Astra. A system card accompanies the release. If you deploy Terra in autonomous agents, that classification is a reminder to keep real permission boundaries in your harness rather than trusting the model to respect them on its own.

GPT-6.1 Terra vs Sol vs Astra: which to pick

Think of it as a tiering decision across the GPT-6.1 lineup:

• GPT-6 Astra — the flagship, still OpenAI's strongest for the hardest computer-use and engineering work, at premium prices.
• GPT-6.1 Sol — the model for complex, multistep coding and agentic work where depth matters more than cost.
• GPT-6.1 Terra — the fast, cheap workhorse for high-volume computer use: forms, invoices, proposals, technical drawings, and video edits, where price-per-task and speed win.

A practical pattern for agent builders: route routine, well-defined computer-use steps to Terra, and escalate only the hardest coding subtasks to Sol or Astra. With Terra priced at half of Sol, that kind of model routing is a real cost lever.

How to access GPT-6.1 Terra

GPT-6.1 Terra is available now to all Plus, Pro, Enterprise, and Edu users in ChatGPT Work and Codex. Developers can call it through the OpenAI API using the model ID gpt-6.1-terra. (The launch was announced by OpenAI on X on October 26.)

Put GPT-6.1 Terra to work in an AI workforce

Terra's whole appeal is doing high-volume computer-use work cheaply — exactly the kind of task you want to hand to an agent rather than do by hand. Eigent is an open-source multi-agent "Cowork" desktop app that turns models like GPT-6.1 Terra into an AI workforce that runs real workflows locally: filling forms, reading documents, automating repetitive desktop tasks, and shipping code. If your team writes software, pair the cheaper tier with an agent that can review GitHub PRs end to end. Download Eigent and route your routine work to the cheapest model that can do it.