阿里巴巴正式推出其最新旗舰AI模型千问Qwen3.8-Max，参数总量达2.4万亿[2]。该公司声称这一模型的性能可与美国领先实验室Anthropic和OpenAI的产品相媲美，同时与国内竞争对手Moonshot AI的Kimi K3相当[1]。在权威榜单中，该模型仅次于Anthropic的Claude系列[2]。
Qwen3.8-Max在多个领域实现了性能升级。该模型能够处理200页财报和100小时长视频[2]，可从空文件夹独立完成十数天的真实项目[2]，并能将法务团队约一周的审阅工作压缩至一小时[2]。模型在编程、办公、长程任务和多模态智能体方面均实现全面提升[2]。目前Qwen3.8的API已在千问AI平台上线，权重将于下周开源[2]。国内用户的定价为每百万Tokens输入12元、输出36元[2]。
Chinese technology giant Alibaba has unveiled Qwen3.8-Max, its most powerful artificial intelligence model to date, featuring 2.4 trillion parameters.[2] The company released the model through a blog post on Monday,[1] claiming its performance rivals leading American AI developers including Anthropic and OpenAI, as well as domestic competitor Moonshot AI's Kimi K3.[1] Alibaba previously stated the model ranks just below Anthropic's Claude system on independent benchmarking leaderboards.[2]
The Qwen3.8-Max represents a significant advancement across multiple capabilities. The model can independently complete complex projects from scratch, compress a legal team's approximately one week of review work into a single hour, and process extensive documents including 200-page financial reports and 100-hour-long videos.[2] Its enhanced performance extends to programming, office work, extended task handling, and multimodal agent applications.[2] The model's API has already launched on the Qwen AI platform, with weights scheduled to open-source the following week.[2] Alibaba priced the service domestically at 12 yuan per million input tokens and 36 yuan per million output tokens.[2]
