Alibaba Group Holding Limited

New York Stock Exchange
Back to all articles
Slightly Bullish +25

Alibaba releases AI model with performance rivalling Anthropic - The Star

🚀 Alibaba Group Holding Ltd has released its latest flagship AI model, Qwen3.8-Max, which rivals global leaders like Anthropic's Fable.

🧠 The new model is built around 2.4 trillion parameters and ranks higher on several benchmarks than Moonshot's recently unveiled Kimi K3.

🌍 Alibaba demonstrated that Qwen3.8-Max delivers comparable or better scores than Anthropic's Fable 5, a model previously subject to US export controls.

📉 This release follows recent market ripples caused by Moonshot's Kimi K3 and DeepSeek's V4 Flash, showing Chinese developers catching up despite computing constraints.

💡 Investors are reassessing the gap between Chinese and Western AI models, with experts noting it is narrowing fast due to these advancements.

📦 Alibaba plans to release the Qwen3.8-Max weights for public download next week, enabling users to customize the technology.

🏃‍♂️ The model excels in autonomous coding and long-horizon execution, having independently executed a 16-day software engineering project in internal testing.

⚡ Efficiency improvements have been made by activating only necessary parts of the system during use to reduce computational costs and latency.

🇨🇳 This launch intensifies the competition among China's AI contenders, including Moonshot, DeepSeek, Z.ai, and ByteDance Ltd.

Bullish Signals
  • Alibaba released Qwen3.8-Max, a flagship AI model built on 2.4 trillion parameters that ranks higher than Moonshot's Kimi K3 on several benchmarks.
  • The model delivers comparable or better scores than Anthropic's Fable 5, demonstrating performance parity with global leaders despite US export controls on advanced capabilities.
  • Internal testing showed the system can independently execute a software engineering project over 16 days, highlighting strong autonomous coding and long-horizon execution capabilities.
  • Efficiency improvements allow the model to activate only necessary parts during use, reducing computational costs and latency.
  • Management intends to release Qwen3.8-Max weights for public download next week, enabling users to customize the technology and expanding its market reach.
Full Analysis
Alibaba Group Holding Ltd has unveiled its latest flagship artificial intelligence model, Qwen3.8-Max, which the company claims delivers performance comparable to global leaders such as Anthropic's Fable. The new model is built around 2.4 trillion parameters and reportedly ranks higher on several benchmarks than Moonshot's recently unveiled Kimi K3. Notably, Alibaba shared results indicating that Qwen3.8-Max achieves scores equal to or better than Anthropic's Fable 5, a cutting-edge model that was previously subject to US export controls due to its advanced capabilities. The release of Qwen3.8-Max follows a period of significant activity in the Chinese AI sector, where Moonshot's Kimi K3 recently caused ripples in stock markets and Silicon Valley by demonstrating rapid catch-up with top models from Anthropic and OpenAI despite constrained computing resources. Vey-Sern Ling, managing director at Union Bancaire Privée, noted that many investors underestimate Chinese AI models due to US chip restrictions or skepticism, arguing that the performance gap is likely much closer and narrowing fast. This sentiment was reinforced by recent expansions in access to DeepSeek's V4 Flash model. Alibaba plans to release the weights for Qwen3.8-Max for public download next week, a move intended to allow users to customize the technology. In internal testing, the model demonstrated strong capabilities in autonomous coding and long-horizon execution, successfully independently executing a software engineering project over 16 days. Additionally, the system has improved efficiency by activating only specific parts when in use, which reduces computational costs and latency. This development marks a significant step in the intensifying race among China's AI contenders, which includes Moonshot, DeepSeek, Z.ai, and ByteDance Ltd.