MiniMax open-sources sparse attention library for NVIDIA Blackwell, M3 weights to launch Friday
KuCoinFlashShare
MiniMax announced the token launch: open-sourcing its high-performance attention library, MiniMax Sparse Attention (MSA), under the MIT license for NVIDIA Blackwell (SM100) GPUs. MSA enables MiniMax-M3 to handle million-token context reasoning, reducing attention computation by 28.4x compared to dense GQA. On H800 GPUs, it achieves 14.2x faster prefill and 7.6x faster decoding speeds. The open-source release includes C++ JIT and CuTe-DSL with multi-precision support. MiniMax-M3 weights are scheduled for release this Friday, marking a significant on-chain update.
Source:Show original
Disclaimer: The information on this page may have been obtained from third parties and does not necessarily reflect the views or opinions of KuCoin. This content is provided for general informational purposes only, without any representation or warranty of any kind, nor shall it be construed as financial or investment advice. KuCoin shall not be liable for any errors or omissions, or for any outcomes resulting from the use of this information.
Investments in digital assets can be risky. Please carefully evaluate the risks of a product and your risk tolerance based on your own financial circumstances. For more information, please refer to our Terms of Use and Risk Disclosure.