NOTICE

vllm.cpp
Copyright 2026 Ettore Di Giacinto

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

This product is a C++ port of vLLM (https://github.com/vllm-project/vllm),
which is licensed under the Apache License, Version 2.0. It incorporates,
ports, and/or vendors code from third-party projects, including:

  - vLLM (Apache-2.0) — the upstream project this is a 1:1 port of.
  - Marlin / GPTQ-Marlin mixed-precision kernels (Apache-2.0), as vendored
    by vLLM under src/vt/cuda/marlin/.
  - NVIDIA CUTLASS (BSD-3-Clause) — used for FP4/FP8 GEMM kernels.
  - xgrammar (Apache-2.0) — C++ grammar core, vendored for structured output.
  - cpp-httplib (MIT) and nlohmann/json (MIT) — header-only dependencies.
  - OpenSSL (Apache-2.0) — a DYNAMIC SYSTEM dependency, not vendored and not
    redistributed here. The build links it (libssl, libcrypto) when
    VLLM_CPP_HF_DOWNLOAD and VLLM_CPP_OPENSSL are on, which is what lets the
    HuggingFace client reach an https endpoint. Its licence travels with the
    system package, and the release manifest records the linked libraries.
  - BoringSSL (a mix of OpenSSL/SSLeay, ISC and MIT licences; see its own
    LICENSE file) — OPTIONAL and STATIC, fetched at configure time only under
    -DVLLM_CPP_BUILD_BORINGSSL=ON and off by default. When that option is used,
    BoringSSL is linked into the binary and its LICENSE must be carried with
    the artifact.
  - google/minja (MIT) - header-only Jinja engine vendored under
    third_party/minja/ (with its LICENSE), used as the chat-template renderer
    in src/vllm/entrypoints/chat_template.cpp.

Each such component remains under its own license; the notices and license
texts of vendored third-party code are retained in their respective source
files and directories.

TRADEMARKS

vllm.cpp is an independent community project. It is NOT affiliated with,
endorsed by, or sponsored by the vLLM project, the PyTorch Foundation, or the
Linux Foundation.

"vLLM" is a mark of its respective owners and is used in this repository in its
descriptive sense only, to identify the upstream project that vllm.cpp ports,
mirrors, and is gated against. No claim of ownership of that mark is made, and
nothing here implies that the vLLM project has reviewed, approved, or endorsed
this software. Section 6 of the Apache License, Version 2.0 grants no trademark
rights, and none are claimed.

Other names and marks referenced here (llama.cpp, SGLang, MLX, CUTLASS,
FlashInfer, Marlin, TRT-LLM, and others) belong to their respective owners and
are likewise used descriptively.