GPT-QModel

license

Apache-2.0(OSI)

source

public

core features withheld

no

The LICENSE file contains the Apache-2.0 text, and the README describes no paid tier, enterprise edition or license-gated build, so the core is ungated. The optional Swordfish Blackwell kernel sources it vendors are licensed under AGPL-3.0-or-later and are only compiled at runtime through JIT, so they are not part of the default build.

• https://raw.githubusercontent.com/ModelCloud/GPTQModel/main/LICENSE recorded 2026-09-30

  LICENSE opens with the header "SPDX-License-Identifier: Apache-2.0" and then runs the verbatim Apache License text, with "Version 2.0, January 2004" and "TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION".

• https://raw.githubusercontent.com/ModelCloud/GPTQModel/main/pyproject.toml recorded 2026-09-30

  pyproject.toml header reads "# SPDX-License-Identifier: Apache-2.0" and the metadata line is license = "Apache-2.0".

• https://raw.githubusercontent.com/ModelCloud/GPTQModel/main/README.md recorded 2026-09-30

  README carries the header "SPDX-License-Identifier: Apache-2.0" and describes "An extensible platform for LLM quantization, validation, and deployment." Its License section says "The optional Swordfish kernel sources vendored under" gptqmodel_ext/swordfish are licensed under AGPL-3.0-or-later and compiled only through JIT. Searched the body for enterprise, license key, paid, commercial and non-commercial: none beyond that optional kernel.

PyPI downloads of `gptqmodel` are the adoption signal here.

• https://pypistats.org/api/packages/gptqmodel/recent recorded 2026-08-18

  last_month downloads = 27,639 for gptqmodel

GPT-QModel applies post-training quantization methods including GPTQ, AWQ, GGUF and FP8, with kernels for five accelerator families and loaders for three serving stacks (Transformers, vLLM and SGLang). That is one optimization family applied broadly across hardware, not the retargetable compilation Apache TVM performs across its whole pipeline.

• https://raw.githubusercontent.com/ModelCloud/GPTQModel/main/README.md recorded 2026-09-30

  README describes "An extensible platform for LLM quantization, validation, and deployment." across "NVIDIA CUDA, AMD ROCm, Huawei Ascend, Intel XPU, and CPU, with Transformers, vLLM, and SGLang."

• https://raw.githubusercontent.com/ModelCloud/GPTQModel/main/README.md recorded 2026-09-30

  README support table lists "Huawei Ascend NPU" with "Native Torch kernels for GPTQ, AWQ, ParoQuant, GGUF, QQQ, and EXL3" and Intel/AMD CPU paths, and says "GPTQ, AWQ, ParoQuant, QQQ, GGUF, FP8, EXL3, GPTAQ, EoRA, GAR, and FOEM".