Alibaba Introduces Qwen3.8: A New Contender in AI
During the World Artificial Intelligence Conference (WAIC) in Shanghai, Alibaba officially announced the next generation of its flagship language model, Qwen3.8. Developers state that this new model boasts an impressive 2.4 trillion parameters and supports multimodal data processing. According to Alibaba’s internal assessments, Qwen3.8’s overall performance is second only to Anthropic’s Claude Fable 5, positioning it as one of the most powerful models currently available.
Availability and Anticipation of Independent Benchmarks
An early version of the model, dubbed Qwen3.8-Max-Preview, is now accessible to subscribers of the Alibaba Token Plan service, as well as on the Qoder and QoderWork platforms. Despite the significant claims regarding its performance, independent benchmark results and comprehensive public comparison tables for Qwen3.8 have yet to be released. This contrasts with recent competitor announcements, where test results often accompany the model’s unveiling.
Alibaba has pledged to open-source the model’s weights soon, which will enable a broader community of researchers and developers to evaluate its capabilities. However, until independent validations confirm its claimed leadership, experts suggest reserving final judgments on Qwen3.8’s standing among top AI models.
I’ve been using Qwen-Max-Preview for a few weeks now through the Token Plan, mostly for content summarization and some basic code generation. The summarization is incredibly fast and surprisingly accurate, even with dense technical papers. However, I’ve noticed it occasionally struggles with very niche programming languages, sometimes generating syntactically incorrect snippets. My tip would be to always double-check its code output, especially for less common frameworks, but for general text tasks, it’s a solid performer.