1 repository on SrcLog
Run Qwen3.5-35B MoE model on RTX 5090 with vLLM using NVFP4 quantization for fast, efficient text generation and extended context length support.