Topics / #nvfp4

Nvfp4 projects on GitHub

Four bit floating point quantization for GPU inference.

As of , the most-starred Nvfp4 projects on GitHub are ZDTaichu5.0-9B (3.1K stars), TypeLLM (960 stars) and GLM-5.3-Flash-EXL3-2x-DGX-Sparks (736 stars). GitHub.Locker tracks 6 Nvfp4 repositories with 5.4K stars combined. The fastest-growing on October 18 was GLM-5.3-Flash-EXL3-2x-DGX-Sparks, with +1 stars in a day.

6Projects
5.4KStars in total
+3Stars in a day
Browse in the catalog →github.com/topics/nvfp4 ↗

Hot right now

Most stars gained in a day · Oct 18

+1

GLM-5.3-Flash-EXL3-2x-DGX-Sparks

MiaAI-Lab/GLM-5.3-Flash-EXL3-2x-DGX-Sparks · ★ 736

A deployment recipe and Docker image that serves the GLM-5.3-Flash model as an OpenAI-compatible vLLM endpoint across 2-4 NVIDIA DGX Spark (GB10) kits, using a custom EXL3/TR3 4-bit quantization of the weights.

+1

dgx-spark-qwen38

hasso5703/dgx-spark-qwen38 · ★ 230

A one-command installer that sets up a hardened, boot-persistent SGLang serving stack for the Qwen3.8 model family on a single NVIDIA DGX Spark (GB10), reaching 50 tok/s greedy median and up to 1M context on the 27B NVFP4 config, with a 176B Flash-Next MoE model also fitting on one box.

+1

v100-skinny

dnv2003/v100-skinny · ★ 175

v100-skinny provides hand-written NVFP4 (W4A16) and FP8 (W8A16) CUDA kernels that let four Tesla V100-SXM2-16GB GPUs run Qwen3.8-27B's published mixed NVFP4/FP8 weights directly, serving the model through an OpenAI-compatible API.

The all-time top

6 projects · by stars

  1. #1 · Taichu-AI/ZDTaichu5.0-9B

    ZDTaichu5.0-9B

    ZDTaichu5.0-9B is a multimodal foundation model that combines a Qwen3.5-9B language backbone with a C-RADIOv4-H vision encoder for general visual understanding, spatial reasoning, embodied-AI research and agentic tool use.

    3.1K
  2. #2 · TypeLLM/TypeLLM

    TypeLLM

    TypeLLM adds type-safe, schema-guaranteed output generation to existing open autoregressive LLMs, without changing their weights or architecture, letting models still think freely while returning typed JSON Schema-constrained answers.

    960
  3. #3 · MiaAI-Lab/GLM-5.3-Flash-EXL3-2x-DGX-Sparks

    GLM-5.3-Flash-EXL3-2x-DGX-Sparks

    A deployment recipe and Docker image that serves the GLM-5.3-Flash model as an OpenAI-compatible vLLM endpoint across 2-4 NVIDIA DGX Spark (GB10) kits, using a custom EXL3/TR3 4-bit quantization of the weights.

    736
  1. 4

    Qwen3.8-27B-DFlash2-EXL3-5.0bpw

    232MiaAI-Lab/Qwen3.8-27B-DFlash2-EXL3-5.0bpw · Python

    A deployment kit for serving Qwen3.8-27B (quantized to EXL3 3.5bpw) through an OpenAI-compatible server, using speculative decoding to speed up generation. It bundles the launcher, server, and config; the model weights live on Hugging Face and inference runs on the team's exllamav3 fork.

  2. 5

    dgx-spark-qwen38

    230hasso5703/dgx-spark-qwen38 · Python

    A one-command installer that sets up a hardened, boot-persistent SGLang serving stack for the Qwen3.8 model family on a single NVIDIA DGX Spark (GB10), reaching 50 tok/s greedy median and up to 1M context on the 27B NVFP4 config, with a 176B Flash-Next MoE model also fitting on one box.

  3. 6

    v100-skinny

    175dnv2003/v100-skinny · Python

    v100-skinny provides hand-written NVFP4 (W4A16) and FP8 (W8A16) CUDA kernels that let four Tesla V100-SXM2-16GB GPUs run Qwen3.8-27B's published mixed NVFP4/FP8 weights directly, serving the model through an OpenAI-compatible API.

Often tagged together