AI Model Registry
Explore and manage local models tested for compatibility. These options are provided as recommendations to run client-side with WebGPU acceleration.
Qwen 2.5 (0.5B)
onnx-community/Qwen2.5-0.5B-Instruct
A highly capable, extremely lightweight model. An excellent balance of size and intelligence for fast browser inference.
Llama 3.2 (1B)
onnx-community/Llama-3.2-1B-Instruct
Meta's official 1B parameter model optimized specifically for edge devices, mobile, and client-side web applications.
TinyLlama (1.1B)
Xenova/TinyLlama-1.1B-Chat-v1.0
A highly popular, compact model trained on 3 trillion tokens. A great baseline for testing web-based LLMs.
Qwen 2.5 Coder (0.5B)
onnx-community/Qwen2.5-Coder-0.5B-Instruct
A version of Qwen 2.5 specifically fine-tuned for writing, understanding, and debugging code in the browser.
Phi-3.5 Mini
onnx-community/Phi-3.5-mini-instruct
Microsoft's highly intelligent edge model. It requires significantly more browser memory but punches well above its weight class.