weightlift
the missing package manager for downloading model weights
Last checked:
What is weightlift?
Weightlift is an in-browser AI model manager that loads, caches, and manages client-side AI models with live progress tracking. Built for JavaScript/TypeScript, it supports transformers.js models and custom model loaders, enabling developers to run inference tasks like vision classification, sentiment analysis, and masked token prediction directly in the browser.
weightlift pricing
Pricing model: Freemium
weightlift pros
- Live progress tracking with byte-level visibility during model loading
- Single ModelManager API for managing multiple AI models
- Built-in support for transformers.js with zero-shot vision, sentiment analysis, and fill-mask tasks
- Extensible via custom ModelDefinition for any runtime or model loader
- Easy npm installation and integration into web applications
weightlift cons
- Client-side execution limited by browser memory and processing constraints
- JavaScript/TypeScript ecosystem only — not available for other languages
- Limited documentation of production-ready examples beyond the playground demo
Frequently asked questions about weightlift
What models does weightlift support?
Weightlift supports transformers.js models out-of-the-box (CLIP, DistilBERT, BERT) and allows custom model loaders via ModelDefinition for any runtime.
How do I install weightlift?
Install via npm: `npm install weightlift`
Can I track model loading progress?
Yes, weightlift provides live progress tracking with dispatch events showing real-time byte counts during model load.
Does weightlift run models entirely in the browser?
Yes, all model inference runs client-side in the browser using the ModelManager instance.