hfviewer

hfviewer is a free web tool that lets you instantly visualize the architecture of any Hugging Face model as an interactive graph. Paste a Hugging Face model URL or repo name and get a clear, zoomable graph of the model's structure in your browser — no local setup, no config files, no notebook required. It is built by Embedl AB as a contribution to the Hugging Face community, and supports a wide range of transformer architectures including LLMs, vision models, multimodal models, and speech models.

• Instant model visualization: Paste a Hugging Face URL or repo ID (e.g., Qwen/Qwen3.5-4B) and the graph opens directly in the browser.
• URL magic: Replace huggingface.co with hfviewer.com in any model URL to jump straight to the visualization.
• Granularity levels: Switch between high-level architecture overview and detailed substructure views to inspect specific blocks and paths.
• Family compare pages: Compare multiple related models (e.g., the Gemma 4 family) side by side with synchronized pan, zoom, and granularity controls.
• Embed in model cards: Press the Embed button to get embeddable code for use in Hugging Face model cards.
• Interactive blog format: Family pages link graph nodes to editorial explanations, enabling a graph-to-text and text-to-graph reading loop.
• Broad model support: Works with decoder-only LLMs, encoder-decoder models, vision transformers, CLIP dual encoders, pose detection models, speech recognition models, and MoE architectures.
• No installation needed: Fully browser-based; the first request may take a few minutes while the server analyzes the model and generates the graph.