アリババ、Meta の挑戦に答えるラップトップ対応 AI モデルを発表

A man walks past the Alibaba logo displayed at its booth during the China International Supply Chain Expo in Beijing, China July 16, 2025.

Florence Lo | Reuters

Meta wants to lead in open-source AI. China’s Alibaba just reminded the U.S. tech giant how competitive the market is.

On Monday, the Chinese tech giant launched an AI model, designed to run on consumer hardware like laptops — a new battleground for AI model developers — and opened the weights of its most powerful model as it looks to maintain its lead in open-source technology.

It comes after Meta announced plans last week to open-source its most powerful AI model and launch new ones designed to run on laptops, in a bid to establish itself as the U.S. alternative to Chinese technology, over frontier labs OpenAI and Anthropic, which use a more proprietary approach.

Launching its new model, Alibaba said Qwen3.8-27B had “excellent capabilities” in handling coding, professional work, research, and long-horizon agentic tasks. It added it matches the performance of another model that is ten times its size.

Alibaba also said it released the weights — calculations and rules that determine how the AI works and behaves — of Qwen3.8 Max, its most powerful model, meaning it can be freely downloaded and run. But the data and methods used to train that model may not be revealed.

Alibaba has established itself as the leader in open-weight AI, with other Chinese companies like DeepSeek and Moonshot also strong players.

watch now

Meta was an early player in open-source AI with its Llama family of models, but was quickly outpaced by Chinese players. Its announcement last week included an open-source family of models called Muse Glimmer, designed to run on a laptop.

“Meta’s own re-embrace of open weights ... was itself a response to two years of Chinese labs ... taking a large share” of the open-weight market, Nick Patience, AI lead at the Futurum Group, told CNBC.

Alibaba looks to cement dominance

The popularity and success of open-weight models is often determined by how many times they’re downloaded and how many developers are using them to build their own products.

Hugging Face, one of the largest repositories from where developers can download open-weight models, said last week that Qwen-based models now account for 151,448 derivatives: times when an open-weight AI model has been downloaded and used. This is 2.6 times Meta’s total footprint, Hugging Face said.

watch now

“The company which can offer the most capable open weights models will move ahead in this race,” Neil Shah, co-founder at Counterpoint Research, told CNBC.

“Alibaba aims to become this undisputed leader, outpacing Meta and eyeing the global market ... as a strong alternative to Silicon Valley frontier-grade deployable models.”

The release of Qwen3.8-27B suggests Alibaba is expecting that more advanced models will be run on the “edge” rather than purely in data centers. Industry experts say that on-device AI could perform faster and be more secure, as it is running on local hardware like a phone or laptop.

Counterpoint Research’s Shah said that “next battleground” for AI models is to succeed running on-device rather than from data centers. Futurum Group’s Patience said Alibaba has built an advantage in several areas, from open-weight models to on-device AI.

“Alibaba has made Qwen the most credible non-US model family to build hardware relationships around, in China and in the open-weight developer community globally,” Patience told CNBC.