🇨🇳中国·AI 新闻·2026年8月18日·来源： SCMP

Alibaba’s lightweight Qwen model takes on larger AI systems from OpenAI, DeepSeek, Zhipu

内容仅提供英文版本

Alibaba Group Holding has unveiled its new lightweight AI model, Qwen3.8-27B, which is designed to compete with larger, more established AI systems from companies like OpenAI. This model, boasting 27 billion parameters, has demonstrated performance comparable to OpenAI’s GPT-5.6 Luna, a model recognized for its cost efficiency. A key advantage of Alibaba’s Qwen3.8-27B is its ability to operate on standard, everyday hardware, making advanced AI more accessible. This development signals a significant step forward for local AI capabilities and momentum within the industry, as highlighted by benchmark firm Artificial Analysis.

Alibaba’s introduction of the Qwen3.8-27B model marks a pivotal moment for Asia’s AI ecosystem, particularly in the context of democratizing advanced AI. By matching the performance of larger models like OpenAI’s GPT-5.6 Luna while being able to run on everyday hardware, Alibaba is directly addressing the high computational costs and infrastructure demands that often hinder broader AI adoption. This move not only enhances Alibaba’s competitive standing against global AI leaders but also empowers a wider range of developers and businesses within Asia to leverage sophisticated AI without prohibitive investments in specialized hardware.

This development underscores a growing trend in the Asian AI market towards efficiency and accessibility. As local AI gains momentum, the ability to deliver high-performance models that are also resource-efficient will be crucial for fostering innovation across various sectors, from startups to established enterprises. It also signals a strategic shift where the focus is not solely on raw parameter count but on practical applicability and cost-effectiveness, potentially accelerating the integration of AI into diverse applications across the region.