--- base_model: - Qwen/Qwen3.5-4B --- Optimized model for ONNX Runtime WebGPU Execution Provider (EP).