Alibaba Just Gave Away Its Best AI Model For Free, Almost Matching Claude and ChatGPT
Alibaba released Qwen3.8-Max as its most capable model yet and plans to open its weights on Hugging Face and ModelScope, its first Max-scale open release. It has 2.4 trillion total parameters, with 95 billion active at a time, emphasizing efficiency and lower deployment cost for smaller teams. Alibaba highlighted long-horizon tasks: the model autonomously built a coding tool over 16 days, reproduced a research paper in five days, and outperformed many human teams in a 24-hour competition. The model supports Claude Code and Codex-style workflows, and Alibaba’s API is compatible with both protocols. Benchmark results are mixed: it trails Anthropic and OpenAI on many text and coding tests, but leads on several multimodal tasks such as documents, video, and spatial reasoning. Its main advantage is cost, reportedly far below top proprietary rivals. The release also signals a strategic shift toward open-weight distribution amid rapid growth in Chinese open models and tightening U.S.-China AI restrictions.
