more than a year since Alibaba vowed to become "more aggressive" in open-source development. “We really feel the power of the open-source community,” said Justin Lin Junyang, the tech lead of Alibaba's Qwen team. “After contributing to the community, the community has also given us a lot of feedback.” The open-source approach gives public access to a program’s source code, allowing third-party software developers to modify or share its design, fix broken links or scale up its capabilities. Open-source technologies have been a huge contributor to China’s tech industry over the past few decades. In February, Alibaba’s updated Qwen family was already powering the world’s top 10 open-source LLMs, according to collaborative machine-learning platform and community Hugging Face. The reported Apple-Alibaba AI partnership stands in stark contrast to the initiatives of major Chinese Android smartphone vendors. Oppo in February said it was integrating R1 into its new foldable handset Find N5, while Honor announced it was using the same DeepSeek model to upgrade its AI agent Yoyo. Vivo had earlier said its smartphone assistant app would incorporate functions backed by R1, such as text creation. Xiaomi, meanwhile, is expected to use its own open-source MiMo AI model in China. For overseas markets, Xiaomi has been working with Google to integrate the Gemini model into its latest smartphones. DeepSeek had generated worldwide attention from late December to January by consecutively releasing two advanced open-source AI models, V3 and R1, which were built at a fraction of the cost and computing power that major tech companies typically require for LLM projects. The latest speculation about DeepSeek-R2 – the successor to the R1 reasoning model – that surfaced in Chinese social media included the product’s imminent launch and the purported new benchmarks it set for cost-efficiency and performance. Meanwhile, the broader LiveBench rankings showed that Qwen3 trailed the world’s leading closed-source AI models – OpenAI’s o3, Google’s Gemini Pro 2.5 and Anthropic’s Claude 3.7, respectively. Microsoft-backed OpenAI’s flagship AI model, o3-mini high, topped the world’s overall ranking of AI models, according to LiveBench. Running o3 costs US$10 for every 1 million tokens. By comparison, running Qwen3 costs 4 yuan (US$0.55) per 1 million tokens, making it more affordable to use. Qwen3’s cost and performance advantages prompted plenty of enterprises to announce their support of Alibaba’s latest AI model on its release. Shenzhen-based Huawei Technologies, along with semiconductor firms Moore Threads, Cambricon Technologies and Hygon Information Technology, have separately announced their support for Qwen3. Cambricon last Tuesday said it successfully optimised Qwen3 to run efficiently on its graphics processing units in response to local AI developers’ demand for China-produced chips. Qwen3 is also being adopted by AI infrastructure companies Hyperbolic and Fireworks.ai on their cloud computing platforms. American semiconductor firms Nvidia and Intel have reportedly started to support Qwen3, too. https://www.scmp.com/tech.. #🇨🇳 ChinA.I. 🤖🧠🦾🤖