vllm.cpp (mudler) GitHub repository

Fields

id

sources/engine-vllm-cpp-github

URL

https://github.com/mudler/vllm.cpp

Publisher

mudler

Kind

repo

Accessed

2026-10-03

notes

mudler/vllm.cpp README: from-scratch C++20 engine claiming vLLM-token-identical output, continuous batching + paged KV + radix cache, CUDA/CPU/Metal/Vulkan backends, GGUF and safetensors, EXL3/MXFP4/NVFP4. Vendor self-reported benchmark numbers; not independently verified.

Notes

mudler/vllm.cpp README: from-scratch C++20 engine claiming vLLM-token-identical output, continuous batching + paged KV + radix cache, CUDA/CPU/Metal/Vulkan backends, GGUF and safetensors, EXL3/MXFP4/NVFP4. Vendor self-reported benchmark numbers; not independently verified.

Cross-references

This record references no other records.