Pinned Loading

1. vllm-sm12x-nvfp4-dflash2 vllm-sm12x-nvfp4-dflash2 Public

   All-NVFP4 vLLM + DFlash2 K7 for Blackwell (SM120/SM121): NVFP4 target/draft/KV, 262K context, optional CPU vision sidecar.

   Python 40 6

2. vllm-sm120-nvfp4-mtp vllm-sm120-nvfp4-mtp Public

   Turnkey vLLM v0.27.1 NVFP4 weights + NVFP4 KV cache + MTP-3 for a single RTX 5090

   Python 7

3. L0xRE L0xRE Public

   L0xRE Releases hub — runtimes and models for low-footprint 27B inference

4. L0xRE-BeeLLama-Low L0xRE-BeeLLama-Low Public

   L0xRE BeeLLama runtime for L0xRE-27b-Low: universal Linux/WSL and Windows packages for RTX 30–50 series; 12 GB VRAM target.

   C++ 3