Qwen2.5-0.5B-Instruct — CoreML (ANE+GPU Optimized)
Converted from Qwen/Qwen2.5-0.5B-Instruct for on-device inference on Apple devices.
| File | Size | Description |
|---|---|---|
model.mlpackage |
302 MB | Monolithic decoder with stateful KV cache (int4) |
- HF-exact match: "The capital of France is Paris." ✅
- iOS 18+ required (MLState API)
See CoreML-LLM for full details.
- Downloads last month
- 15