MicroLLM Lab runs seven tiny language models entirely in your browser

MicroLLM Lab – Try 7 tiny LLM's in the browser

MicroLLM Lab lets you run, benchmark, and compare seven small language models — from a 26M MiniMind2 up to a 362M SmolLM2 — directly in the browser via WebGPU, with Q4 quantization shrinking them to 15–216 MB. Everything stays on-device through IndexedDB caching, with no servers or accounts. The lab also generates shareable benchmark certificates and supports custom JavaScript evals.