MiniMax 开源高性能 MSA kernel 库,并预告 M3 模型权重将于本周五发布
多源视角
day-0 and already on @FireworksAI_HQ with blazing fast inference<br><br>long-horizon agents, full-repo understanding, multimodal coding all in one model<br><br>Try M3 today on Fire…
查看这条来源M3 is live on @telnyx Inference on day-0<br><br>go build with Telnyx and M3 today<hr style="border:0;border-top:1px solid #80808030;margin:12px 0;"><div class="rsshub-quote">Telnyx…
查看这条来源Run M3 locally today with @UnslothAI<hr style="border:0;border-top:1px solid #80808030;margin:12px 0;"><div class="rsshub-quote">Unsloth AI: MiniMax M3 can now be run locally!🔥<br…
查看这条来源M3 is now live on @parasail_io 🚀<hr style="border:0;border-top:1px solid #80808030;margin:12px 0;"><div class="rsshub-quote">Parasail: Minimax M3 is live on Parasail, day zero. It…
查看这条来源M3 open weight just dropped and it's live on @Modular cloud on day zero <br><br>with up to a 1M-context and MSA architecture kernel-to-cloud optimization is exactly what M3 needs<b…
查看这条来源RT Novita AI<br>🤗 MiniMax M3 from @MiniMax_AI is now live on @huggingface — supported by Novita.<br><br>Open weights.<br>~428B total parameters.<br>~23B activated parameters.<br>B…
查看这条来源MiniMax M3, Open-Weight, Now On Hugging Face , with only ~428B parameters and ~23B activated parameters<br><br>Weights:<br>https://huggingface.co/MiniMaxAI/MiniMax-M3<br>MiniMax Sp…
查看这条来源RT @RyanLeeMiniMax: On the M3 license — thanks for the feedback on M2.7<br>You told us prior-approval-for-any-commercial-use was too much. We…
查看这条来源RT LMSYS Org<br>🎉 SGLang has Day-0 support for MiniMax-M3 from @MiniMax_AI, a native-multimodal MoE reasoning model of ~428B total params (~23B active), 60 layers, 1M context acro…
查看这条来源MiniMax M3, Open-Weight, Now On Hugging Face<br><br>Weights:<br>https://huggingface.co/MiniMaxAI/MiniMax-M3<br>MiniMax Sparse Attention:<br>https://huggingface.co/papers/2606.13392…
查看这条来源RT Mike Gannotti<br>Early morning testing over coffee of minimax-m3:cloud for audio/video/images ingestion, writing and coding capabilities via @ollama cloud. @MiniMax_AI<br><img w…
查看这条来源Excited to cohost this one with @cysic_xyz <br><br>With $5,000 dollars in prizes and 80% off M3, there is about to be some incredible projects built.<br><br>Now is your chance, wha…
查看这条来源AI 摘要
MiniMax 宣布开源高性能 MSA kernel 库,并预告 M3 模型权重将于本周五发布。MSA 库为稀疏注意力提供底层高性能实现,开发者可通过 GitHub 获取代码与论文。此举降低了稀疏注意力优化门槛,同时 M3 开放权重将让研究者能够本地部署该模型进行后续研究与开发。
推荐理由高信息密度,值得细读
原文
无法获取正文,可打开下方原始链接查看。
讨论
暂无评论。