GPT-6.1 Terra is OpenAI's newest answer to a question every business using artificial intelligence keeps asking: how do you get capable AI help for everyday computer work without paying flagship prices? Launched on Monday 26 October 2026, GPT-6.1 Terra is a fast, efficient model built for the tasks that fill a normal workday — reading what is on screen, clicking through apps, filling in forms, drafting documents and editing video — and OpenAI says it does that work at half the price of the model it replaces. In a week already crowded with AI news, GPT-6.1 Terra is trending per news coverage because it changes the cost calculation for AI agents, not just the capability leaderboard.
OpenAI describes GPT-6.1 Terra as a significant step up from GPT-5.6 Terra, the previous version of its mid-tier model. The pitch is deliberately practical rather than spectacular. Where flagship models are sold on solving the hardest research or coding problems, GPT-6.1 Terra is sold on volume: more tasks completed, faster, for less money, inside ChatGPT Work and Codex and through OpenAI's API. That positioning matters because most real-world AI use is not a single brilliant answer — it is hundreds of small jobs, repeated daily, where speed and price decide whether automation is worth doing at all.
What Is GPT-6.1 Terra?
GPT-6.1 Terra sits in the middle of OpenAI's current model family. OpenAI now organises its models into four main classes, from lowest price and capability to highest: Luna, Terra, Sol and Astra. Terra is the workhorse tier: smarter and more capable than the cheapest option, but far cheaper than the Sol and Astra models built for complex coding, professional work and in-depth research. If Sol and Astra are the specialists you call in for the hardest jobs, GPT-6.1 Terra is the colleague who handles the everyday workload.
In OpenAI's own words, GPT-6.1 Terra is "a fast, efficient model that combines detailed visual reasoning and precise computer use alongside clearer, more natural writing." That combination is the key to understanding it. Visual reasoning means the model can interpret what is actually displayed on screen — a technical drawing, a chart, a form, a spreadsheet — rather than only processing typed text. Precise computer use means it can then act on that understanding: navigating software, completing multi-step processes and producing a finished result. OpenAI lists examples including drafting proposals, interpreting technical drawings, filling out forms and editing videos.
For ordinary users, the practical difference is in ChatGPT Work and Codex, where GPT-6.1 Terra is now available to Plus, Pro, Enterprise and Edu users. OpenAI says a user's usage allowance goes further with GPT-6.1 Terra than with GPT-5.6 Terra — and far further than with GPT-6.1 Sol or GPT-6 Astra. Developers can also call it directly through the OpenAI API under the model name gpt-6.1-terra, and Amazon Web Services has made it generally available through Amazon Bedrock, where it accepts text and image input with a context window of up to 1 million tokens.
GPT-6.1 Terra Pricing: Half the Cost of Its Predecessor
The headline number in the GPT-6.1 Terra launch is price. OpenAI has set standard API pricing at $1 per million input tokens and $5 per million output tokens, with cached input priced at $0.05 per million tokens. That is 50% lower than GPT-5.6 Terra, the model it replaces, and 50% lower than GPT-6.1 Sol. For comparison, OpenAI positions Terra against Anthropic's Claude Opus 5.5, which it lists at $8.82 per comparable unit in its benchmark chart — making Terra roughly one-tenth the price of that rival flagship on a per-token basis.
Token pricing can feel abstract, so it helps to translate it. A million input tokens is roughly 750,000 words of text fed into the model. At $1 per million, feeding a large report, a set of drawings or a long email thread into GPT-6.1 Terra costs a fraction of a cent per page in most cases, while the $5 output price applies to the text the model writes back. Batch and Flex processing cost half the standard rates again, according to OpenAI's model page, which makes high-volume, non-urgent work cheaper still. The model also carries a 1,050,000-token context window and a knowledge cutoff of 1 June 2026, meaning it can work with very large documents in a single session.
Why does a 50% price cut matter more than a small capability bump? Because AI agents only make economic sense when the cost of running them is lower than the value of the time they save. A model that is slightly less brilliant but three times faster and half the price can automate jobs that a flagship model would be too expensive to touch — processing invoices, triaging support tickets, updating records across apps, or drafting routine reports. OpenAI's bet with GPT-6.1 Terra is that the next wave of AI adoption will be won on that arithmetic, not on benchmark records.
How Fast and Capable Is GPT-6.1 Terra?
OpenAI's performance claims for GPT-6.1 Terra rest mainly on one internal test: Agents' Last Exam, an evaluation of how well an agent can use software to complete complex, multi-step work tasks. On that test, GPT-6.1 Terra scored 15 percentage points higher than GPT-5.6 Terra at medium reasoning effort, while costing about one-third as much. OpenAI also reports that Terra performs on par with its flagship GPT-6 Astra on that exam, at one-tenth of Astra's price per token, and completes tasks more than three times faster on average than both GPT-6 Astra and GPT-6.1 Sol.
A second internal measure tells a similar story about reliability. On OpenAI's benchmark of prompts designed to induce errors, 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 percentage points to 4.3%, nearly matching the rates for GPT-6 Astra and GPT-6.1 Sol. In plain terms, the new model is far less likely to go wrong on routine tasks even when it is not thinking its hardest, which is exactly the mode in which most everyday work happens.
These figures deserve one honest caveat, and it applies to every vendor launch: they are OpenAI's own measurements, not independent tests. Outside analysts have already noted that GPT-6.1 Terra does not beat Anthropic's most advanced models outright — one comparison put it 3.2 percentage points behind Claude Opus 5.5 on the same exam, while costing far less. The fair reading is therefore not that Terra is the best AI model in the world, but that it closes much of the gap to the best models on computer-use tasks while costing a fraction of the price. Real-world proof will come from users running it on their own work over the coming weeks, not from launch-day charts.
GPT-6.1 Terra vs GPT-6 Astra and GPT-6.1 Sol
Choosing between OpenAI's models now comes down to the kind of work you are doing. GPT-6 Astra remains the flagship for the hardest problems — deep research, complex reasoning and the most demanding professional tasks. GPT-6.1 Sol is the step below it, aimed at complex coding and professional work. GPT-6.1 Terra is the everyday tier: detailed visual reasoning, precise computer use and clear writing, delivered quickly and cheaply.
OpenAI's own guidance draws the line clearly. For in-depth research or complex coding, it says, choose GPT-6.1 Sol or GPT-6 Astra. For everything else — the communication and computer-based tasks that fill the workday — GPT-6.1 Terra is the intended default, and its usage allowance in ChatGPT Work and Codex stretches further than either of its bigger siblings. That is a notable shift in advice: instead of always reaching for the biggest model, OpenAI is now telling users that the mid-tier model is the right tool for most jobs, and that using it saves both money and time.
The competitive context explains why OpenAI is pushing this message now. The launch landed on the same day Anthropic filed its public IPO prospectus, a coincidence that put the two rivals' strategies side by side: Anthropic asking public markets to value its frontier models, OpenAI cutting prices to widen everyday use. OpenAI sets Terra against Claude Opus 5.5 throughout its announcement, framing the contest as capability-per-dollar rather than capability alone. Whether that framing wins customers will depend on whether Terra's computer-use skills hold up outside OpenAI's test suite — but the pricing pressure it creates is real for every AI provider.
Who Gets GPT-6.1 Terra and How to Use It
GPT-6.1 Terra is rolling out now. ChatGPT Plus, Pro, Enterprise and Edu users can use it in ChatGPT Work and Codex, where it appears as a model option for everyday tasks. Developers can access it through the OpenAI API as gpt-6.1-terra, through both the Responses and Chat Completions interfaces, and through Amazon Bedrock for teams that run their AI workloads on AWS. One practical detail for developers: the model supports a reasoning effort setting of "none", which then allows temperature and top-p controls that were blocked on GPT-6.1 Sol — useful for teams that want faster, more predictable outputs for routine jobs.
For a small business or a busy professional, the most useful way to think about GPT-6.1 Terra is as an automation budget. Jobs that were previously too expensive to hand to an AI agent — because each attempt cost too much, or took too long — become worth trying at half the price and three times the speed. That includes reading and summarising long documents, extracting data from forms and drawings, drafting standard letters and proposals, and carrying out repetitive multi-app workflows. The model will still make mistakes, as every AI model does, so high-stakes work still needs a human check. But the cost of that check is now much lower, and the range of jobs worth automating is much wider.
What GPT-6.1 Terra is not, at least not yet, is a replacement for human judgement or for specialist tools. It cannot verify facts beyond its training data on its own, its knowledge cutoff is 1 June 2026, and its launch benchmarks say nothing about how it performs on any one company's private systems. The sensible approach is to start with low-risk, high-volume tasks, measure the error rate on your own work, and expand from there. Used that way, GPT-6.1 Terra looks less like a revolution and more like something arguably more important: AI assistance becoming cheap and fast enough to be ordinary.
Related reading: Anthropic IPO 2026: Public S-1 Filed, Ticker ANT Named, Price Still Unset · SpaceX $40 Billion AI Financing: Filing Confirms Apollo Lease — But the Deal Is Not Closed Yet · Humanoid Robots 2026: Factories First, Homes Later — What Toyota Is Actually Doing
Sources: OpenAI announcement via the OpenAI Developer Community, 26 October 2026 · BleepingComputer, "OpenAI launches GPT-6.1 Terra with 3x faster AI agents, 50% lower pricing" · 9to5Mac, "OpenAI upgrades ChatGPT and Codex with new GPT-6.1 Terra model" · XDA Developers, "GPT-6.1 Terra brings more advanced computer agent use at a lower cost" · MarketWatch live coverage of the Anthropic IPO filing and OpenAI launch, 26 October 2026 · Amazon Web Services via Oossa, "OpenAI GPT-6.1 Terra launches on Amazon Bedrock".