<div id="readability-page-1">Semiconductor designer Moore Threads Technology has achieved full-stack compatibility between its flagship MTT S5000 graphics processing unit (GPU) and Alibaba Cloud’s Qwen3.5-series artificial intelligence models, as China’s tech self-reliance efforts gather pace. Beijing-based Moore Threads, founded by former Nvidia executive James Zhang Jianzhong, on Thursday said the firm’s flagship AI chip was now compatible with the three new models under the latest Qwen series – Qwen3.5-35B-A3B, Qwen3.5-122B-A10B and Qwen3.5-27B. Alibaba Cloud is the AI and cloud computing services unit of Hangzhou-based Alibaba Group Holding, owner of the South China Morning Post. The announcement comes a day after Alibaba Cloud released its Qwen 3.5 medium model series, part of the Qwen 3.5 family unveiled earlier this month. The company had earlier touted how the series performed in certain benchmarks with leading models from OpenAI, Anthropic and Google DeepMind. Moore Threads’ initiative reflects how Chinese chip developers are rushing to support the country’s top-performing AI models. These GPU designers – including MetaX Integrated Circuits, Biren Technology and Enflame – are in a heated race to fill the void left by Nvidia in the domestic market. Nvidia does not know if and when its much-anticipated H200 chips would be allowed into China, the company said on Thursday. That showed the ongoing regulatory uncertainty from Beijing regarding GPU imports, which Chinese “big tech” companies are keen to acquire for their AI development projects.  In a separate move, AI start-up DeepSeek has given Chinese chip designers early access to its coming V4 model for performance optimisation, something it has not done with US chip suppliers, according to a report on Wednesday by Reuters, citing sources familiar with the matter. Huawei Technologies was among the domestic suppliers with access to the new DeepSeek model, the report said, adding that this pre-release engagement was not standard industry practice. DeepSeek and Huawei did not immediately respond to requests for comment on Friday. For Moore Threads, which made a splash on the capital market through its Shanghai listing in December, it has stepped up efforts to make its GPUs compatible with the latest Chinese AI models on the market. Earlier this month, the company announced “day zero” chip compatibility with Zhipu AI’s GLM-5 model, which was launched ahead of China’s Spring Festival holiday. As part of the chip designer’s latest initiatives, developers now can develop new models using the company’s native programming language based on its MUSA – for Meta-computing Unified System Architecture – ecosystem. That would make the development process much easier by lowering the difficulty of transforming from Nvidia’s CUDA – for Compute Unified Device Architecture – platform. These platforms provide the software layer that lets developers harness the power of GPUs for applications.</div>

YINN

512480

159995

<p>Moore Threads Technology has announced full-stack compatibility of its MTT S5000 GPU with Alibaba Cloud's Qwen3.5-series AI models, enhancing China's tech self-reliance. This compatibility includes three new Qwen models, following Alibaba Cloud's recent release. The move highlights the competitive landscape among Chinese chip developers, as they seek to fill the gap left by Nvidia amid regulatory uncertainties. Additionally, AI start-up DeepSeek has provided early access to its V4 model for performance optimization to domestic chip designers, including Huawei. Moore Threads aims to simplify GPU development with its native programming language, easing the transition from Nvidia's platform.</p>

Moore Threads’ flagship AI chip compatible with latest Alibaba models