FlashMLA

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/deepseek-ai/FlashMLA/blob/main/LICENSE recorded 2026-09-02

  LICENSE body is the verbatim MIT text under a DeepSeek copyright line.

• https://api.github.com/repos/deepseek-ai/FlashMLA recorded 2026-09-02

  Repo metadata for flashmla's canonical repository - private false, archived false, fork false.

• https://github.com/deepseek-ai/FlashMLA/blob/main/README.md recorded 2026-09-02

  README describes FlashMLA as DeepSeek's library of optimized attention kernels powering the DeepSeek-V3 and DeepSeek-V3.2-Exp models, with no paid tier, enterprise edition or license-gated build beside the published source.

12,895 GitHub stars, which lands in the >10K stars band of the stars scale, level 3. The `flash-mla` name on PyPI holds a single 1.0.0.dev0 release with no summary or project URL, which is not DeepSeek's distribution; the README installs from source with `pip install -v .`, so no download count exists and the stars scale applies with its cap of 3.

• https://api.github.com/repos/deepseek-ai/FlashMLA recorded 2026-09-02

  Repo metadata - stargazers_count = 12,895 - for deepseek-ai/FlashMLA.

One band below flash-attention: both are attention kernel libraries, but FlashMLA's kernels are purpose-built for DeepSeek's own multi-head latent attention and sparse attention variants rather than serving as a general-purpose attention primitive across model architectures.

• https://github.com/deepseek-ai/FlashMLA/blob/main/README.md recorded 2026-09-02

  README describes FlashMLA as DeepSeek's library of optimized attention kernels powering the DeepSeek-V3 and DeepSeek-V3.2-Exp models, with no paid tier, enterprise edition or license-gated build beside the published source.