Hugging Face’s AI agent uses a Linux virtual machine to complete tasks autonomously.
Photo Credit: Hugging Face
Hugging Face’s latest AI agent is powered by the Qwen2-VL-72B AI model
Hugging Face released a demo of an artificial intelligence (AI) agent on Tuesday that can complete various web-based tasks. Dubbed Open Computer Agent, the tool is available for free, and anyone can go to the website to access the agent. Since it can perform tasks on web browsers, it can access websites such as Google Search, Maps, and even ticket booking platforms to complete tasks autonomously. Notably, this AI agent is part of Hugging Face's smolagents library, which was introduced in January.
In a post on X (formerly known as Twitter), Aymeric Roucher, Project Lead - Agents at Hugging Face, announced the release of Open Computer Agent. As the name suggests, it is an open-source computer-based agent that can autonomously complete a wide range of tasks. The agent has been provided with a Linux virtual machine and several apps, and the Mozilla Firefox web browser.
Roucher explained that the AI agent is powered by Qwen2-VL-72B, a vision language AI model that can identify elements in an image by their coordinates. The model allows the agentic tool to analyse whatever is on the screen, take action, and then move to the next step. Notably, the agentic capability in the model is being added by Hugging Face's smolagents tool.
We're launching Computer Use in smolagents! 🥳
— m_ric (@AymericRoucher) May 6, 2025
-> As vision models become more capable, they become able to power complex agentic workflows. Especially Qwen-VL models, that support built-in grounding, i.e. ability to locate any element in an image by its coordinates, thus to… pic.twitter.com/mI8MuWZkIS
It is a free-to-use AI agent, meaning anyone can click here to go to the Open Computer Agent website and try out the tool. Users can ask the computer use agent to find directions to a location, and it will open the browser, go to Google Maps, type in the origin and destination location, and start navigation to show the result.
Gadgets 360 staff members were able to test out the AI agent. While it functions as advertised, we found it to be quite slow in completing tasks. It also makes mistakes or fails to perform a task when the prompt is complex. Additionally, since it is a cloud-based free tool, there is a long queue, and visitors might have to wait several minutes before the agent starts working on the task.
Get your daily dose of tech news, reviews, and insights, in under 80 characters on Gadgets 360 Turbo. Connect with fellow tech lovers on our Forum. Follow us on X, Facebook, WhatsApp, Threads and Google News for instant updates. Catch all the action on our YouTube channel.
Supermoon and Geminid Meteor Shower 2025 Set to Peak Soon: How to See It