luzimu/WebGenAgent-LM-7B-SFT

TEXT GENERATIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:7.6BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Aug 30, 2025License:mitArchitecture:Transformer Open Weights Featherless Exclusive Cold

The luzimu/WebGenAgent-LM-7B-SFT is a 7.6 billion parameter language model developed by luzimu, specifically fine-tuned for autonomous website generation from natural language instructions. It leverages multi-level feedback, including visual and functional assessments, and step-level reinforcement learning to iteratively refine website codebases. This model excels at creating interactive and functional websites by understanding complex requirements and continuously improving its output.

Loading preview...

WebGenAgent-LM-7B-SFT Overview

This model, developed by luzimu, is a 7.6 billion parameter language model specifically designed for autonomous website generation from natural language instructions. It is part of the WebGen-Agent framework, which combines advanced language models with specialized training to create functional and visually appealing websites.

Key Capabilities

  • Iterative Website Generation: Generates and refines website codebases through a multi-step process.
  • Multi-Level Feedback Integration: Utilizes both visual feedback (via a Visual Language Model) and functional feedback (via a GUI-agent) to assess and improve generated websites.
  • Step-Level Reinforcement Learning: Employs a Step-GRPO (Gradient Policy Reinforcement Optimization) approach, using screenshot scores and GUI-agent scores as dense, reliable process supervision.
  • Natural Language Understanding: Interprets natural language instructions for both appearance and functional requirements of a website.

Good For

  • Developers and researchers focused on automated web development and code generation from high-level descriptions.
  • Applications requiring interactive website creation with continuous refinement based on visual and functional criteria.
  • Exploring advanced agentic AI systems that integrate multiple feedback loops for complex task completion.