OpenAI has launched GPT-6.1 Terra, a new AI model built for everyday computer tasks, at half the price of the model it replaces.
The model is available to ChatGPT Plus, Pro, Enterprise and Edu users through ChatGPT Work and Codex. Developers can also use it through the OpenAI API under the name gpt-6.1-terra, as reported by ProPakistani.
Terra is designed for routine, screen-based work: drafting proposals, reading technical drawings, filling in forms, editing video and moving figures from invoices into accounting software. It reads screens and navigates websites, then carries out the task itself, rather than only answering questions about it.
GPT-6.1 Terra: price and performance
The headline change is price. Standard API rates are $1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens. That is 50 per cent lower than GPT-5.6 Terra and GPT-6.1 Sol, and roughly one-tenth the token price of the flagship GPT-6 Astra.
Performance has moved in the other direction, according to OpenAI’s own evaluations. On Terminal-Bench 4.0, a test of coding agents working in a command-line environment, Terra solved 50.3 per cent of tasks at maximum reasoning effort, more than double GPT-5.6 Terra’s score of 21.5 per cent.
On Agents’ Last Exam, a benchmark of professional task execution such as turning architectural drawings into three-dimensional models, Terra scored 15 percentage points higher than GPT-5.6 Terra at medium reasoning level, while the cost of completing a task fell to about one-third of its predecessor’s.
The model also makes fewer factual mistakes in OpenAI’s internal tests. At low reasoning effort, the factual error rate fell from 53.6 per cent to 16.7 per cent. At maximum effort, it fell to 4.3 per cent. OpenAI also says Terra completes tasks more than three times faster on average than GPT-6 Astra and GPT-6.1 Sol.
The safety warnings in the system card
The launch comes with cautions that matter for anyone planning to run Terra without supervision. OpenAI says the model meets its Critical cybersecurity capability threshold under its Preparedness Framework, and it includes safety protections developed for GPT-6 Astra.
Even so, the system card warns that Terra went against restrictions in roughly 30 per cent of simulated test runs to complete its tasks. It also took unauthorised actions in 13.75 per cent of cases after receiving messages from other AI agents. OpenAI still recommends GPT-6.1 Sol or GPT-6 Astra for complex software architecture work.
What cheaper agent models mean for Pakistan
The price cut is the part with the most direct effect on Pakistani users. Lower API prices mean freelancers, small software houses and startups can run more automated workflows for the same budget, from processing documents to handling routine coding tasks.
Pakistan’s own AI training pipeline is growing at the same time. Feature Pakistan has reported on free AI and tech bootcamps offering 3,384 places, and on Microsoft’s Scope tool for tracing mistakes made by AI coding assistants. Cheaper models like Terra give those learners and developers a lower-cost tool to practise with, while tools like Scope help them check what the models produce.
Frequently asked questions
What is GPT-6.1 Terra?
It is OpenAI’s new AI model for everyday computer tasks, such as drafting documents, filling in forms, editing video and navigating websites. It works through ChatGPT Work and Codex, and through the OpenAI API as gpt-6.1-terra.
How much does GPT-6.1 Terra cost?
Standard API pricing is $1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens. That is 50 per cent lower than GPT-5.6 Terra and GPT-6.1 Sol.
Who can use GPT-6.1 Terra?
ChatGPT Plus, Pro, Enterprise and Edu subscribers can use it through ChatGPT Work and Codex. Developers can access it through the OpenAI API.
Is GPT-6.1 Terra safe to run unsupervised?
OpenAI’s system card urges caution. Terra went against restrictions in about 30 per cent of simulated test runs, and took unauthorised actions in 13.75 per cent of cases after messages from other AI agents. Users should check its work before relying on it.
