Vetted AI models that run on your device. Every file is an approved parameter file served once from models.skillsafe.ai — download it in one SkillSafe app and the next app already has it. Nothing you enter is uploaded to run it, and it never costs a credit. Models we convert ourselves are also published on Hugging Face — the same bytes with their provenance, for use outside SkillSafe. Missing one? Request it — or see how the registry works.
Feature extraction model for transformers.js (fp32). Runs on WASM or WebGPU; 86.2 MB in one file, downloaded once and cached for every SkillSafe app.
Image classification model for transformers.js (q8). Runs on WebGPU or WASM; 24.6 MB in one file, downloaded once and cached for every SkillSafe app.
Image segmentation model for onnxruntime-web (int8 dynamic-quantized). Runs on WASM; 13.1 MB in 2 files, downloaded once and cached for every SkillSafe app.
Image segmentation model for onnxruntime-web (fp16). Runs on WebGPU or WASM; 19.8 MB in 2 files, downloaded once and cached for every SkillSafe app.
Image classification model for transformers.js (q8). Runs on WebGPU or WASM; 84.2 MB in one file, downloaded once and cached for every SkillSafe app.
Automatic speech recognition model for transformers.js (fp32). Runs on WebGPU or WASM; 278 MB in 2 files, downloaded once and cached for every SkillSafe app.
Object detection model for onnxruntime-web. Runs on WebGPU or WASM; 34.2 MB in one file, downloaded once and cached for every SkillSafe app.
Object detection model for onnxruntime-web. Runs on WebGPU or WASM; 3.5 MB in one file, downloaded once and cached for every SkillSafe app.
Image super resolution model for onnxruntime-web (fp32). Runs on WebGPU or WASM; 4.6 MB in one file, downloaded once and cached for every SkillSafe app.
File type detection model for onnxruntime-web (fp32). Runs on WASM; 3.0 MB in one file, downloaded once and cached for every SkillSafe app.
Text recognition model for onnxruntime-web (fp32). Runs on WASM or WebGPU; 4.3 MB in one file, downloaded once and cached for every SkillSafe app.
Text detection model for onnxruntime-web (fp32). Runs on WASM or WebGPU; 1.7 MB in one file, downloaded once and cached for every SkillSafe app.
Background removal model for onnxruntime-web (fp32). Runs on WASM or WebGPU; 4.4 MB in one file, downloaded once and cached for every SkillSafe app.
Portrait matting model for onnxruntime-web (fp16). Runs on WASM or WebGPU; 12.4 MB in one file, downloaded once and cached for every SkillSafe app.
Text generation model for web-llm (q4f16_1). Runs on WebGPU; 2055 MB in 84 files, downloaded once and cached for every SkillSafe app.
Text generation library model for web-llm. Runs on WebGPU; 5.2 MB in one file, downloaded once and cached for every SkillSafe app.
Image to text model for transformers.js (q8). Runs on WebGPU or WASM; 479 MB in 2 files, downloaded once and cached for every SkillSafe app.
Image to text model for transformers.js (fp32). Runs on WebGPU or WASM; 914 MB in 2 files, downloaded once and cached for every SkillSafe app.
Image classification model for transformers.js (fp32). Runs on WebGPU or WASM; 97.4 MB in one file, downloaded once and cached for every SkillSafe app.
Translation model for transformers.js (q8). Runs on WASM or WebGPU; 244 MB in 2 files, downloaded once and cached for every SkillSafe app.
Translation model for transformers.js (fp32). Runs on WASM or WebGPU; 404 MB in 2 files, downloaded once and cached for every SkillSafe app.
Image classification model for transformers.js (fp32). Runs on WebGPU or WASM; 330 MB in one file, downloaded once and cached for every SkillSafe app.
Translation model for transformers.js (q8). Runs on WASM or WebGPU; 241 MB in 2 files, downloaded once and cached for every SkillSafe app.
Image super resolution model for transformers.js (fp32). Runs on WebGPU or WASM; 51.9 MB in one file, downloaded once and cached for every SkillSafe app.
The registry is empty. Request the first model.