BREAKING
Technology

OpenAI’s GPT-6.1 Terra brings AI-powered computer control

OpenAI has released GPT-6.1 Terra, an artificial intelligence model. It’s not just for answering questions. This model can see what’s on a computer screen. It can also interact with apps and programs. It was introduced on October 26 2026. Because of this it can do things that usually need a person to do them. Like clicking buttons typing or navigating menus. It’s like having a helper that sees and acts. This could change how we use computers every day.

The launch marks another step toward AI systems that can actively work with AI software than just generate AI text or provide AI answers. OpenAI CEO Sam Altman announced the model on X (Twitter) saying, “6.1 Terra is here! It is good at a lot of things. OpenAI is especially excited, for this new era of computer use.”

AI that can interact with apps

GPT-6.1 Terra is designed to work with a computer in a way that feels like a person. GPT-6.1 Terra can read what is, on the screen click buttons, type words and jump from one app to another while following a list of tasks.

This capability could be especially helpful for businesses that handle digital tasks. The model might for example move data from invoices into accounting systems fill out online forms, draft business proposals or turn architectural blueprints into 3D models. The goal is to let AI manage workflows, with very little help from people.

Faster and more affordable

OpenAI says GPT-6.1 Terra delivers a significant performance improvement compared with the previous GPT-5.6 Terra. According to the company, the new model completes tasks roughly three times faster on average than more powerful models such as GPT-6.1 Sol and GPT-6 Astra.

Cost is another major part of the upgrade. OpenAI has reduced standard API pricing by 50 per cent compared with the previous version. Developers now pay $1 per million input tokens and $5 per million output tokens, potentially making the model more attractive for businesses that need to run automated workflows at scale.

Better accuracy and coding capabilities

The company has also focused on reducing errors. In its internal testing, GPT-6.1 Terra’s error rate on simple tasks fell from 53.6 per cent with the older model to 16.7 per cent. At maximum reasoning effort, the error rate dropped further to 4.3 per cent.

The model has also made significant gains in coding, achieving more than twice the performance of its predecessor on complex terminal benchmarks. GPT-6.1 Terra is therefore already being made available to programmers through platforms such as GitHub Copilot.

Where Terra fits in

OpenAI’s current model lineup includes Luna, Terra, Sol and Astra. Luna is positioned as a lightweight model, while Sol and Astra are intended for more demanding workloads. Terra sits between these options, offering a balance of speed, capability and cost.

Following recent updates to Sol and Luna, GPT-6.1 Terra brings the Terra series into the GPT-6.1 generation. The model is now available through ChatGPT Plus, Pro, Enterprise and Edu subscriptions in ChatGPT Work and Codex, as well as through OpenAI’s API and Amazon Bedrock.