DigiNews

Tech Watch by Johan Denoyer

← Back to articles

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

Quality: 8/10 Relevance: 9/10

Summary

MicroLLM Lab is an in-browser environment that lets you load, run, benchmark, and compare small language models (SLMs) with hardware-accelerated WebGPU. It emphasizes on-device privacy, zero server costs, and no accounts, using 4-bit quantization to fit 25M–360M parameter models into 50–84 MB of browser memory and providing a shareable performance certificate.

🚀 Service construit par Johan Denoyer