OpenAI has released GPT-6.1 Terra, a new mid-tier AI model built for everyday computer work. Launched on October 26, 2026, it replaces GPT-5.6 Terra and is designed to read screens, fill in forms, draft documents and handle the repetitive office tasks that eat up an afternoon — at half the API price of its bigger sibling, GPT-6.1 Sol.
What GPT-6.1 Terra is
Terra sits in the middle of OpenAI's model lineup, between the cheapest Luna tier and the more powerful Sol and Astra models. OpenAI describes it as a fast, efficient model that pairs sharper visual reasoning and precise computer use with clearer, more natural writing. It is available in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu users, and through the API under the model id gpt-6.1-terra.
The price cut that matters
API pricing is $1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens. OpenAI says that is 50% lower than GPT-5.6 Terra and half the standard rates of GPT-6.1 Sol. Requests above 272,000 input tokens are charged at higher long-context rates. The context window is about 1.05 million tokens, with up to 128,000 tokens of output, and the model's knowledge cutoff is June 1, 2026, according to OpenAI's developer documentation.
What it can actually do
OpenAI's examples are deliberately unglamorous: reading incoming invoices and entering them into accounting software, filling in forms, drafting proposals, reading technical drawings and editing video. That is the point. Terra is pitched at high-volume, well-scoped work rather than research breakthroughs — the kind of tasks where a model clicking, typing and moving between apps saves real hours.
How it compares
On OpenAI's own benchmarks, Terra scores 50.3% on Terminal-Bench 4.0 at maximum reasoning effort, up from 21.5% for GPT-5.6 Terra, and its factual-error rate on the company's difficult-prompts test falls from 53.6% to 16.7% at low effort and 4.3% at maximum effort. These are company-reported figures, not independent tests. Independent benchmarking firm Artificial Analysis puts Terra's Intelligence Index at 48.1 with a cost of $0.40 per task, against $1.40 per task for GPT-5.6 Terra — cheaper, but not a frontier model: Claude Opus 5.5, Kimi K3.1 Pro and Gemini 4 Argon all score higher in the same index. The comparison is familiar from this month's other big AI money stories, from the record funding rounds covered in Meta's $95 billion AI funding blitz to the IPO filing examined in Anthropic's record-breaking IPO — the industry is now competing as much on cost per task as on raw capability.
Why cheaper computer use changes the math
When a model costs half as much per task, letting it loose on thousands of invoices, forms or support tickets stops being an experiment and starts being a budget line. That is why the agent angle matters: the same week has seen AI agents move from demos into real workflows, as covered in Walmart letting AI agents shop for you, and rival open-weight models pushing prices down further, as with Moonshot's Kimi K3.1 Pro launch. Terra's bet is that the winning model for office work is not the smartest one, but the one cheap and reliable enough to run all day.
What to watch
Rollout depends on plan, region and workspace settings, so not every user will see Terra on day one. OpenAI also notes that Sol and Astra remain the better choices for the most complex coding tasks. For everyone else, the question is practical: how much of your workday is clicking through the same screens, and what happens when that work costs a dollar per task instead of eight?
GPT-6.1 Terra will not win benchmark charts, and it is not trying to. It is OpenAI's attempt to make computer-using AI cheap enough to become boring — in the best sense. If it works, the most important AI model of the year might be the one you barely notice. Would you trust an AI to handle your invoices and forms, or do you still want a human checking every click? Tell us in the comments.
Comments
Post a Comment