Ternlight – 7 MB embedding model that runs in browser (WASM)
7 MB embedding model. Runs entirely in your browser. No API calls, no latency, no vendor lock-in.
Why it matters
Ternlight demonstrates a emerging shift toward edge-deployed, ultra-lightweight AI models that eliminate infrastructure dependency. For builders, this means faster prototyping and deployment without cloud costs; for enterprises, it's a new model for on-device inference at scale.
The key facts
11 to knowModel size: 7 MB
Deployment: Browser-native (WASM)
Community interest: 213 points on HN with 49 comments
No external API dependency
Embedding model (semantic search/retrieval capability)
Published July 2026
Runtime: WebAssembly (WASM) in-browser execution
Capability: Embedding generation without server dependency
Community validation: 213 points, 49 comments on Hacker News (strong developer interest)
Live demo available at
Published: July 6, 2026
Go to the source
Hacker Newsternlight-demo.vercel.app
Publisher excerpt: Article URL: Comments URL: Points: 213 # Comments: 49