AI BriefingGUI Automation VLM 'Holo1' Released
Key point
Holo1, an open-source VLM family specialized for GUI automation, and the web agent Surfer-H have been released.
Details
H Company has launched Holo1, an Action VLM family optimized for GUI understanding and precise localization, along with WebClick, a multimodal localization benchmark.
Holo1 consists of Holo1-3B and Holo1-7B models, with the 7B model achieving 76.2% UI localization accuracy, the best among small models. These models are based on the Qwen2.5-VL architecture and are fully compatible with the Transformers library.
Surfer-H, a web-native agent built on Holo1, operates the browser just like a real user and consists of three independent modules:
- Policy model: Plans and drives the agent's actions
- Localizer model: Visually understands the UI for precise interaction
- Validator model: Confirms whether the task has been completed
Surfer-H demonstrated excellent cost efficiency on the WebVoyager benchmark. It achieved 92.2% accuracy on real-world web tasks while lowering the cost per task to as low as $0.13, enhancing its practicality.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Whiteboard Launches Open-Source VS Code Fork for Human-AI Collaborative Software Design
OpenComputer Hosts 'Ship Video' Agent Using Claude Opus 5.5 to Generate Explainer Videos from Code