Alibaba has released a new AI model under its Qwen brand, specifically engineered to run on consumer-grade hardware such as laptops, while also opening up the model weights of its most powerful system, Qwen3.8 Max. These moves escalate the competition between the Chinese tech giant and Meta in the open-weight AI space.
Meta has long aimed to lead the open-source AI race, but Alibaba just delivered a sharp reminder of how fiercely contested that field has become. On Monday, the Chinese company unveiled an AI large language model tailored for edge devices like laptops, a segment that has quickly emerged as a new battleground for AI developers. At the same time, Alibaba released the weights for its flagship model, seeking to cement its dominance in open-source technology.
The announcement comes just a week after Meta revealed plans to open-source its own most advanced AI model and introduce several new models capable of running locally on laptops. Meta's strategy is to position itself as the leading American alternative to closed-source powerhouses like OpenAI and Anthropic, while also countering China's growing AI capabilities.
Alibaba's newly launched Qwen3.8-27B delivers standout performance in coding, professional office tasks, scientific research, and long-horizon intelligent agent operations, with the company claiming it can match the capabilities of models ten times its size. Additionally, Alibaba has made the weights of its flagship Qwen3.8 Max available for download. These weights contain the computational logic and behavioral rules that drive the AI model, allowing developers to freely download, deploy, and run it. However, the training data and methodologies used to build the model remain undisclosed.
Alibaba has established itself as a frontrunner in the open-weight AI segment, though domestic competitors like DeepSeek and Moonshot AI also remain formidable players. Meta entered the open-source AI space early with its Llama series, but Chinese firms quickly caught up and surpassed it in adoption. Last week, Meta rolled out its Muse Glimmer open-source model series, designed specifically for local execution on laptops.
Nick Peshin, head of AI at Futurum Group, commented on the shift: "Meta's renewed push into the open-weight space is itself a direct response to Chinese labs capturing a significant share of the open-source market over the past two years."
Alibaba moves to solidify its leading position
Success in the open-weight model arena largely hinges on download volumes and the size of the developer ecosystem building on top of those models. Data from Hugging Face, the world's largest open-source model repository, showed last week that the number of models derived from Qwen has reached 151,448, a figure 2.6 times larger than the entire ecosystem built around Meta's models.
Neil Shah, co-founder of Counterpoint Research, noted: "Whoever offers the most capable open-weight model will gain a decisive edge in this race. Alibaba is aiming to become the undisputed leader, surpassing Meta and targeting a global audience as a compelling alternative to the top deployable models coming out of Silicon Valley."
The release of Qwen3.8-27B signals Alibaba's belief that high-performance AI models will increasingly run at the edge rather than solely in data centers. Industry experts point out that on-device AI running locally on phones and laptops offers faster response times and enhanced security.
Shah added that the "next battleground" for large language models lies in local, on-device deployment rather than reliance on cloud data centers. Peshin echoed this view, stating that Alibaba has already built advantages across both open-weight models and edge AI.
"Qwen has emerged as the most compelling non-American model ecosystem, earning strong recognition both through hardware partnerships in China and within the global open-source developer community," Peshin said.
Comments