webSLM Demo — Base + RAG vs Fine-tuned SLM

100% in-browser inference via WebGPU • No server required • Powered by WebLLM & MLC-LLM

Option 1 Base Model — RAG panel
Option 2 webSLM-Medical-0.5B — fine-tuned SLM panel
Custom compiled webSLM model — loads from HuggingFace (VishalMysore/WebSLM-Custom-MLC).
Try:
Option 1
Base + RAG
General model • Documents retrieved and injected into context
Initialising…
Option 2
Fine-tuned webSLM
Domain-specialized model • No retrieval • Behaviour baked in during training
Initialising…