Alibaba and Huawei Announce Kimi K3 Support to Strengthen China’s AI Ecosystem

China’s artificial intelligence industry has witnessed another major development after Alibaba Group and Huawei announced Kimi K3 support shortly after Moonshot AI made the full weights of its flagship open-weight model publicly available on 27 July. The rapid response from two of China’s largest technology companies highlights their commitment to strengthening the country’s AI infrastructure and expanding enterprise access to advanced artificial intelligence.

The announcements also demonstrate how quickly China’s technology ecosystem can respond to new AI innovations. By offering immediate compatibility through cloud platforms and AI hardware, Alibaba and Huawei are helping organisations deploy next-generation AI models without significant delays.

Alibaba and Huawei Roll Out Kimi K3 Support

Alibaba Cloud confirmed that its Zhenwu M890 supernode achieved Day-0 compatibility, allowing customers to run the model as soon as the open weights were released. In addition, Alibaba revealed that the model would be available through its Qwen AI platform and Bailian Model Studio API, providing developers and enterprises with flexible options for integrating advanced AI into their applications.

Huawei followed with its own announcement, confirming that the Ascend 950 series processors and Atlas A3 AI platform are fully compatible with the new model. According to Huawei, the model’s 896-expert Mixture-of-Experts (MoE) architecture is well aligned with the company’s ultra-node computing design, enabling efficient execution of large-scale AI workloads.

Further expanding deployment capabilities, Qujing Tech successfully adapted Kimi K3 for inference on Huawei’s Ascend Atlas A3 platform using the open-source SGLang engine. This achievement provides organisations with additional options for deploying AI applications on Huawei-powered infrastructure.

The swift Kimi K3 support announced across cloud services, AI hardware, and deployment platforms reflects China’s coordinated effort to accelerate the adoption of domestically developed AI technologies.

Why Kimi K3 Is Receiving Global Attention

Developed by Moonshot AI, Kimi K3 is one of the largest open-weight language models released to date.

The model is built on a 2.8-trillion-parameter Mixture-of-Experts architecture consisting of 896 experts, with only a portion of the parameters activated during inference. This design delivers high performance while improving computational efficiency. It also features a one-million-token context window, allowing users to process lengthy documents, research material, code repositories, and complex datasets within a single session.

Moonshot AI has also introduced several architectural enhancements, including Kimi Delta Attention, Attention Residuals, Stable LatentMoE, native vision capabilities, and reinforcement learning improvements. Together, these technologies enhance reasoning, software development, mathematical problem-solving, and AI agent performance.

Benchmark Results Highlight Strong Performance

The model has attracted widespread industry attention following its benchmark performance.

According to the Artificial Analysis Intelligence Index, Kimi K3 ranks third globally, behind Anthropic’s Claude Fable 5 and OpenAI’s GPT-5.6 Sol. It also secured the top position in frontend code generation, demonstrating strong capabilities for software engineering tasks.

Following the public release of the model weights, Hugging Face CEO Clem Delangue stated that the model became one of the platform’s fastest-growing releases, reaching the trending charts within just 30 minutes.

Competition in the AI Industry Continues to Grow

The release has intensified competition across the global AI landscape.

Earlier this month, Alibaba previewed its Qwen3.8-Max model during the World AI Conference in Shanghai, further demonstrating the company’s ongoing investment in advanced AI technologies. At the same time, reports suggest that US policymakers are evaluating additional restrictions on Chinese open-weight AI models, highlighting the growing strategic importance of AI development.

Many industry analysts have described the launch as another significant milestone for China’s AI industry, showing that domestic companies continue to narrow the technology gap through advances in model architecture, cloud infrastructure, and deployment capabilities.

Open-Weight AI Gains Wider Adoption

The popularity of the model quickly exceeded Moonshot AI’s available computing capacity, leading the company to temporarily pause new subscriptions shortly after launch.

Following the open-weight release, several global AI infrastructure providers, including Nebius, Baseten, and Fireworks AI, introduced deployment support for the model. Developer platforms such as Cursor and Cognition’s Devin also integrated it into their AI-powered tools, making it more accessible to software engineers and enterprise users.

This expanding Kimi K3 support across Chinese and international platforms illustrates the increasing demand for open-weight AI models that offer greater flexibility and deployment freedom.

Key Developments

Announcement	Details
Alibaba Cloud	Day-0 compatibility on Zhenwu M890 supernode
Enterprise Availability	Integrated into Qwen AI platform and Bailian Model Studio API
Huawei	Full support across Ascend 950 series and Atlas A3
Deployment Partner	Qujing Tech enabled inference using the SGLang engine
Model Size	2.8-trillion-parameter Mixture-of-Experts model
AI Ranking	Third on the Artificial Analysis Intelligence Index
Coding Performance	First in frontend code generation benchmarks
Global Adoption	Supported by Hugging Face, Nebius, Baseten, Fireworks AI, Cursor, and Devin

Why This Development Matters

The growing Kimi K3 support ecosystem demonstrates that China’s AI companies are investing beyond model development. Cloud providers, chip manufacturers, software platforms, and infrastructure companies are working together to create an environment that enables faster deployment of advanced AI technologies.

For enterprises, this means easier access to powerful AI capabilities without the need for extensive infrastructure investments. Developers benefit from greater flexibility when building AI applications, while organisations can choose from a broader range of deployment options suited to their operational needs.

Conclusion

The rapid rollout of Kimi K3 support by Alibaba, Huawei, and multiple AI infrastructure providers marks an important milestone in China’s evolving AI strategy. Immediate cloud compatibility, hardware optimisation, and broad deployment support have enabled the model to move quickly from research into practical enterprise use. Coupled with strong benchmark results and growing international adoption, these developments reinforce China’s expanding role in the global AI industry. As demand for open-weight models continues to rise, the collaborative approach taken by technology companies is expected to accelerate innovation and shape the next generation of enterprise artificial intelligence.