Alibaba Launches 2.4 Trillion Parameter AI Model With No Independent Benchmark Scores
Alibaba released Qwen3.8-Max-Preview on July 19, 2026, claiming it ranks second globally among frontier AI models, behind only Anthropic's Claude Fable 5. As of July 21, independent evaluators Artificial Analysis and benchlm.ai each show zero published benchmarks for the model, leaving the ranking unverifiable. A single limited coding test by Trilogy AI found the model scoring 80 out of 100 against Moonshot's Kimi K3, which scored 83, suggesting the claim of second place is not yet supported by reproducible evidence.