Moore Threads quickly completed comprehensive adaptation for the Qwen3.5 model.

```

Today, Moore Threads announced that it has completed comprehensive adaptation of Alibaba's latest large model Qwen3.5 on its flagship AI training and inference all-in-one full-featured GPU MTT S5000. This support fully demonstrates the maturity and completeness of the Moore Threads MUSA ecosystem, allowing developers to efficiently deploy and optimize models using the MUSA C programming language and Triton-MUSA toolchain.

During the adaptation of Qwen3.5, Moore Threads verified two core capabilities of the MUSA ecosystem: Native MUSA C support enables developers to directly use MUSA C for kernel development, greatly reducing the migration threshold from the CUDA ecosystem; Deep compatibility with Triton-MUSA allows developers to write high-performance operators using familiar Triton syntax and seamlessly run them on Moore Threads’ full-featured GPU via the Triton-MUSA backend.

For the hybrid attention mechanism used in the Qwen3.5 multimodal model, Moore Threads achieved native optimization. Based on the muDNN computing library and the MATE open source operator library, Moore Threads provides efficient support for long sequence processing in the hybrid attention mechanism, successfully enabling high-performance inference of the model on the MTT S5000. This accomplishment not only once again verifies the wide adaptability and efficient support capability of domestic full-featured GPU compute platforms for cutting-edge large models, but also demonstrates the significant effects of coordinated software-hardware optimization.

From GLM-5 to MiniMax M2.5, and now to Qwen3.5, Moore Threads has consistently kept up with and adapted to domestic top-tier large models at high speed. This normalized agile response mechanism is not only rooted in the MUSA architecture's seamless compatibility with mainstream AI ecosystems and its continuously optimized toolchain support, but also marks that the domestic computing foundation now has full-chain support capabilities, from model adaptation to efficient deployment. In the future, Moore Threads will continue to deepen its MUSA technology ecosystem, and with a more solid and user-friendly domestic computing foundation, assist more cutting-edge large models in landing applications at the earliest opportunity, accelerating the prosperity of the domestic computing ecosystem.

Source: Moore Threads Official Account

Risk Warning and DisclaimerThe market involves risks, and investment requires caution. This article does not constitute personal investment advice, nor does it take into account individual users’ specific investment objectives, financial conditions, or needs. Users should consider whether any opinions, viewpoints, or conclusions in this article are suitable for their particular circumstances. Investment is at your own risk. ```