Dev Tools · 2h ago
Ternlight: 7MB AI Embedding Model Runs Entirely in Browser via WASM
Ternlight is a 7MB embedding model that runs in the browser using WebAssembly, enabling on-device AI without server calls. It allows developers to generate text embeddings locally, reducing latency and privacy concerns. The model is small enough for quick downloads and efficient inference on consumer hardware.
Meridian48 take
While 7MB is impressively small, the real test is whether its embedding quality competes with larger models for practical search or RAG tasks.
embedding-modelwebassembly