Alibaba Launches Qwen3.8-Max: Top-Tier Performance at Bargain Prices
Alibaba/Qwen
Anthropic
Alibaba has released its new flagship large language model Qwen3.8-Max, which claims top-tier performance on the Arena leaderboard, rivaling Anthropic's Claude models. With a total of 2.4 trillion parameters, it excels in coding, professional work, long-horizon tasks, and multimodal agent capabilities, all at a significantly lower price point than competitors.
Alibaba has introduced Qwen3.8-Max, a new generation foundational large language model with 2.4 trillion parameters. The model quickly climbed to the top tier on the Arena leaderboard, with performance approaching Anthropic's Claude series and even surpassing Claude Opus 5 and Fable 5 in coding. Pricing is aggressive: domestic input costs 12 CNY per million tokens, output 36 CNY, and cache hits only 1.5 CNY; international prices are just 40% of Opus 5 for input and 24% for output. Early testers have praised its coding and content creation abilities, noting its affordability. In hands-on tests, the model successfully built a fully functional AI news webpage from a single product requirement document, performed cross-chapter analysis of a technical paper, and generated a detailed timeline for a 70-minute podcast. Qwen3.8-Max is available on the Qianwen AI platform and integrated into the new 'Qianwen Office' agent product, continuing Alibaba's open-source legacy with over 400 models released since 2023.
Source: QbitAI 量子位 —
original