Infrastructure / Compilers & Model Optimization

FlashQLA

FlashQLA is a linear attention kernel library from the Qwen team, built on TileLang, applying operator fusion to the forward and backward passes of Gated DeltaNet (GDN) chunked prefill. It serves as a backend for the flash-linear-attention library's GDN operators through the standard FLA API.

Verified 2026-09-02 via the GitHub API and the LICENSE body.

Openness

5 high confidence

5.0

license: MIT(OSI)
source: public
core-gated: ungated

MIT license body confirmed. The repository is public and unarchived and builds the whole product, and the README describes no paid tier, enterprise edition or license-gated build beside it, so source is public and the core ungated.

https://github.com/QwenLM/FlashQLA/blob/main/LICENSE recorded 2026-09-02
LICENSE body is the verbatim MIT text under an Alibaba/Qwen copyright line.
https://api.github.com/repos/QwenLM/FlashQLA recorded 2026-09-02
Repo metadata for flashqla's canonical repository - private false, archived false, fork false.
https://github.com/QwenLM/FlashQLA/blob/main/README.md recorded 2026-09-02
README describes FlashQLA as a high-performance linear attention kernel library built on TileLang serving as a flash-linear-attention GDN backend, with no paid tier, enterprise edition or license-gated build beside the published source.

Adoption

1 high confidence

1.0

2,326 PyPI downloads of `flash-qla` in the trailing 30 days, which lands in the <10K band of the software usage scale, level 1. `pip install flash-qla` is the README's install path and the package description links back to QwenLM/FlashQLA.

https://pypistats.org/api/packages/flash-qla/recent recorded 2026-09-02
last_month downloads = 2,326 for flash-qla
https://pypi.org/pypi/flash-qla/json recorded 2026-09-02
PyPI metadata for flash-qla 0.1.2 - summary 'FlashQLA: Fused TileLang kernels for Linear Attention', description links https://github.com/QwenLM/FlashQLA, requires tilelang.

Capability

2 medium confidence

2.0

Banded on the category feature matrix as a narrow kernel set: one linear-attention variant's forward/backward path, built to plug into one upstream API. Placed two bands below the tensorrt rung, the same level as flashkda.

https://github.com/QwenLM/FlashQLA/blob/main/README.md recorded 2026-09-02
README describes FlashQLA as a high-performance linear attention kernel library built on TileLang serving as a flash-linear-attention GDN backend, with no paid tier, enterprise edition or license-gated build beside the published source.

Verified 2026-09-02