WebVIA-Agent-f32-GGUF

WebVIA-Agent is part of the WebVIA framework, representing the first agentic vision-language model built for interactive, verifiable UI-to-Code generation on web interfaces. Based on a GLM-4.1V-9B backbone, WebVIA-Agent navigates websites to capture multi-state UI screenshots and interprets both image and DOM tree inputs to identify which interactive components require actions, factoring in previous interactions and occlusions to avoid redundant operations. The broader framework includes UI2Code for generating executable HTML/CSS/JavaScript and a validation module to ensure the generated UIs behave as intended, achieving more stable and accurate UI exploration than general-purpose agents like Gemini-2.5-Pro. WebVIA-Agent thus enables reliable automation and tool-support for realistic, dynamic web environments by not only modeling static layouts, but also verifying and automating interactive workflows.

Model Files

Quants Usage