GStars
    deepseek-ai

    deepseek-ai/FlashMLA

    FlashMLA: Efficient Multi-head Latent Attention Kernels

    developer-tools
    C++
    MIT
    13.3K stars
    1.3K forks
    13.3K GitHub watchers
    Updated Nov 5, 2026
    View on GitHub

    Build with Backblaze B2

    SDKs, agent skills, IDE extensions, and reference pipelines from Backblaze Labs. All open source.

    Explore Backblaze Labs

    Loading star history...

    Where it stands

    Rank, growth pace and release cadence relative to everything GStars tracks.

    The project lists 19 contributors on GitHub.

    Use Cases & Benefits

    • Implements optimized multi-head latent attention kernels for efficient sparse and dense attention in transformer models.
    • Delivers high-performance attention computation with up to 660 TFLOPS on NVIDIA GPUs using FP8 quantization and CUDA 12.8+.
    • Use for accelerating token-level sparse attention in large language model decoding with FP8 KV cache support.
    • Use for high-throughput dense multi-head attention prefill and decoding on SM90 and SM100 GPU architectures.
    • Use for integrating efficient attention kernels into PyTorch 2.0+ workflows requiring state-of-the-art GPU performance.

    Should you depend on it?

    License terms, project hygiene and risks, from GitHub metadata and GStars' snapshots.

    License

    MIT

    Permissive: commercial use, modification and redistribution are allowed with attribution.

    Community files

    • README: present

    Watch-outs

    • No GitHub releases; pin a commit or tag instead of a release.

    General guidance from repository metadata, not legal advice. How we score

    From the README

    FlashMLA is DeepSeek's library of optimized attention kernels, powering the DeepSeek-V3 and DeepSeek-V3.2-Exp models. This repository contains the following implementations:

    These kernels power DeepSeek Sparse Attention (DSA), as introduced in this paper.

    Excerpt from the deepseek-ai/FlashMLA README on GitHub (MIT license). © its authors.Read the full README on GitHub

    Daily snapshots

    Stars gained per day, and how forks and open issues moved, from GStars' daily snapshots

    Forks

    1.3K+148

    Open issues

    147+27

    Compared with peers in Developer Tools

    Fastest-growing repositories GStars tracks in this cohort for the week of Nov 1, by 7-day star growth

    View category →
    deepseek-ai/FlashMLA compared with Developer Tools repositories by stars, 7-day star growth, health score and last release
    RepositoryStars+7dHealthLast release
    stablyai/orca116.8K+6.6K40today
    affaan-m/ECC297.9K+5.2K6335 days ago
    DietrichGebert/ponytail185.3K+5.1K607 days ago
    herdrdev/herdr48.3K+1.7K5922 days ago
    The-PR-Agent/pr-agent15.1K+1.5K6411 days ago
    deepseek-ai/FlashMLA13.3K—17—
    Category median116.8K+5.1K60—
    Compare FlashMLA with orca, ECC, ponytail →

    Also by deepseek-ai

    Other repositories from this maintainer that GStars tracks

    Discover Repositories

    Search across tracked repositories by name or description