OpenAI Launches GPT-6.1 Terra at Half the API Price
OpenAI has launched GPT-6.1 Terra, a faster computer-use AI model at half the API price of its predecessor, with better accuracy for everyday professional work.
OpenAI has launched GPT-6.1 Terra, a new AI model built for everyday professional work. It replaces GPT-5.6 Terra and focuses on computer use: understanding what is on screen, reading documents and drawings, and carrying out multi-step tasks across applications. The headline change is the price. Standard API rates are half those of its predecessor and of GPT-6.1 Sol, OpenAI's higher-tier model.
What OpenAI announced
GPT-6.1 Terra is a reasoning model that sits in the middle of OpenAI's current line-up, between the lower-cost GPT-6 Luna and the more powerful GPT-6.1 Sol and GPT-6 Astra. OpenAI describes it as a fast, efficient model that combines detailed visual reasoning and precise computer use with clearer, more natural writing. It is aimed at teams that need to automate large volumes of routine work, rather than at researchers chasing the highest possible score on hard reasoning tests. Typical examples OpenAI gives include drafting proposals, interpreting technical drawings, filling out forms, editing videos, processing invoices, and entering details into accounting software.
What it can do better
On Agents' Last Exam, a benchmark of professional tasks, OpenAI reports that GPT-6.1 Terra at maximum effort scored 60.0 percent at a cost of 1.01 dollars per task. That compares with 50.7 percent for GPT-5.6 Terra at 3.58 dollars per task, and 59.3 percent for the flagship GPT-6 Astra at 7.24 dollars per task. In other words, OpenAI's own numbers put Terra slightly ahead of Astra on this test at roughly one-seventh of the cost per task. Coding also improves sharply on OpenAI's reported tests. On Terminal-Bench 4.0, a test of solving command-line coding tasks, GPT-6.1 Terra scored 50.3 percent at maximum effort, against 21.5 percent for GPT-5.6 Terra. On the OSWorld 2.1 benchmark, which tests navigating desktop applications, it scored 74.2 percent at maximum effort, up from 57.1 percent for its predecessor. OpenAI also says the model makes fewer errors on prompts designed to trip it up. At low reasoning effort, the error rate fell from 53.6 percent with GPT-5.6 Terra to 16.7 percent with GPT-6.1 Terra, and at maximum effort it fell to 4.3 percent. These are OpenAI's own evaluations, so independent testing will matter before teams rely on them for critical work.
The price in plain terms
API prices are charged per million tokens, the small chunks of text a model processes. GPT-6.1 Terra costs 1 dollar per million input tokens, 0.05 dollars per million cached input tokens, and 5 dollars per million output tokens. That is exactly half the standard rates of GPT-6.1 Sol, which costs 2 dollars, 0.10 dollars, and 10 dollars on the same measures, and far below GPT-6 Astra at 10 dollars, 1 dollar, and 50 dollars. The practical point is volume. A team running thousands of invoice checks, document reviews, or form-filling jobs a day pays far less per job than with a flagship model, even if each job still needs a human to check important results. But cheaper tokens alone do not make a workflow cheap: the time spent reviewing outputs, fixing mistakes, and connecting the model to company software still counts.
Who can use it, and where
GPT-6.1 Terra is available to Plus, Pro, Enterprise, and Edu users in ChatGPT Work and Codex. Developers can access it through the OpenAI API using the model name gpt-6.1-terra. On safety, OpenAI says Terra meets the Critical threshold for cybersecurity under its own Preparedness Framework and keeps the safeguards developed for GPT-6 Astra. It also reports that the model is more likely than GPT-5.6 Terra to stay within the scope of a user's request and to disclose its limitations or blockers, and that it resisted prompt-injection attempts designed to divert it from the assigned task in testing.
What to watch
The honest test for Terra is not a benchmark chart but real office work. Analysts quoted in the coverage advise enterprises to test it on their own workloads and measure task completion, accuracy, time per job, cost per completed workflow, and how much human review is still needed. If those numbers hold up, Terra could become the default workhorse model for routine AI automation. If they do not, the lower price will not save a workflow that needs constant correction. For now, the launch sends a clear signal about where the AI market is heading: the race is no longer only about the smartest model, but about the cheapest model that is good enough for the job.
Sources
- Computerworld - https://www.computerworld.com/article/4241023/openai-bets-on-ai-economics-with-lower-cost-gpt-6-1-terra-2.html - BusinessToday - https://www.businesstoday.in/technology/news/story/openai-launches-gpt-6-1-terra-with-advanced-computer-use-capabilities-at-half-the-api-price-563959-2026-10-27