OpenAI released GPT-6.1 Terra on October 26, 2026, and it's the company's new fast, lower-cost model for computer use, visual reasoning, coding and everyday office work. It replaces GPT-5.6 Terra, sits below GPT-6.1 Sol and GPT-6 Astra in OpenAI's lineup, and comes with API pricing that's half of what GPT-5.6 Terra and GPT-6.1 Sol cost. If you build with AI agents or use ChatGPT for work, this is the model most likely to show up in your daily tools over the next few weeks.
Here's what Terra is, what it costs, how OpenAI says it performs, and how to try it today.
What GPT-6.1 Terra is built for
OpenAI pitches Terra as the model you reach for when you want an agent to get real work done quickly and cheaply. According to gHacks' summary of the launch, the examples OpenAI gave include drafting proposals, reading technical drawings, filling out forms, editing videos and entering invoice details into accounting software. Those are computer-use tasks, where the model looks at a screen and clicks, types and moves between apps the way a person would.
That focus fits the bigger shift toward agents that carry out tasks on your behalf. If you're new to the idea, our explainer on what agentic AI is covers the basics, and our roundup of AI agents you can text shows what that looks like in a messaging app.
GPT-6.1 Terra pricing
Terra's API pricing is the headline for developers and startups. Per DataCamp's breakdown, which checked OpenAI's pricing on October 26, the model costs:
$1 per million input tokens
$0.05 per million cached input tokens
$5 per million output tokens
The API model ID is gpt-6.1-terra. For a team running agents that read a lot of screenshots and documents, cutting the per-token price in half matters a lot, since computer-use workflows can burn through input tokens fast. OpenAI hasn't published a context window figure for Terra in the coverage we reviewed, so check the model page in your API dashboard before planning around long documents.
Benchmarks, according to OpenAI
All of the performance numbers so far come from OpenAI's own testing, so treat them as claims until independent evaluations catch up. With that caveat, the reported results are big jumps over the previous Terra model.
On Terminal-Bench 4.0, a test of coding and command-line tasks, OpenAI says Terra solves 50.3% of tasks at maximum effort, up from 21.5% for GPT-5.6 Terra. On Agents' Last Exam, Terra scores 15 percentage points higher than GPT-5.6 Terra at medium reasoning effort, at about a third of the cost, and OpenAI says it performs on par with GPT-6 Astra while finishing tasks more than 3 times faster than both Astra and GPT-6.1 Sol.
Factual accuracy also improved on OpenAI's internal error-inducing prompt test. At low effort, the error rate dropped from 53.6% with GPT-5.6 Terra to 16.7% with Terra, and at maximum effort it fell to 4.3%, close to the larger models. DataCamp also notes Terra beat GPT-5.6 Terra across reasoning settings on MMMU-Pro, a multimodal reasoning test. OpenAI still recommends GPT-6.1 Sol or GPT-6 Astra for the most complex coding work, so think of Terra as the efficient everyday default and save the bigger models for the hardest problems.
How to try GPT-6.1 Terra
In ChatGPT, Terra is available to Plus, Pro, Enterprise and Edu subscribers inside ChatGPT Work and Codex, per gHacks and DataCamp. OpenAI hasn't said whether it will come to the free tier, and rollouts can take a little while to reach every account. On those paid plans, OpenAI says the same usage allowance covers more tasks with Terra than with GPT-5.6 Terra, and substantially more than with Sol or Astra.
Developers can call it through the API using the gpt-6.1-terra model ID. A good first test is a task you already run with another model, such as summarizing a batch of support tickets or filling in a spreadsheet from PDFs, so you can compare cost and quality side by side. If you're deciding between providers more broadly, our comparison of ChatGPT vs Claude vs Gemini is a useful starting point, and our guide to open-weight vs closed AI models covers when running your own model makes more sense.
Safety notes from the launch
OpenAI says Terra meets the "Critical" threshold for cybersecurity under its Preparedness Framework and keeps the safeguards developed for GPT-6 Astra. The company also reports stronger resistance to prompt injection than GPT-5.6 Terra, which matters for any agent that reads web pages or emails, since hidden instructions in that content are one of the main ways computer-use agents get tricked. Details are in OpenAI's system card.
What it means for everyday users
For most people, Terra's arrival means faster, cheaper agents inside the tools they already pay for. Expect the apps you use for email, spreadsheets and scheduling to switch defaults as developers chase the lower price, and expect a wave of startups building computer-use products on top of it. Our guide to AI workflows for work has practical ideas for putting a model like this to use without handing it the keys to everything.
If you like comparing notes on new models as they drop, the AI Trends Tribe is a good place to trade prompts, benchmark takes and real-world tests after each launch. And if you'd rather talk shop in person, subscribe to the Tribe Events calendar on Luma for founder workshops, trivia and meetups around the Bay.
