Alibaba's new Qwen3.8 Max AI model claims to match unsupervised working skills touted by rivals like Anthropic's Claude, but independent benchmarks are lacking, raising questions about its performance.
The release highlights the intensifying competition between Chinese and US AI developers, particularly in the wake of US restrictions on advanced chip exports to China. Alibaba's efforts to showcase its AI capabilities suggest a drive to demonstrate parity or leadership in the field.
Alibaba has launched its latest AI model, Qwen3.8 Max, claiming it can perform complex tasks unsupervised for extended periods, rivaling capabilities previously touted by Anthropic's Claude. The Qwen team stated the model can design computer chips and rewrite research papers without human oversight. In one test, the model reportedly worked for approximately five days to rebuild and improve a math research paper's experiment. In another, it competed in a 24-hour online contest against 526 human teams, outperforming all but 68.
However, the performance claims have been met with skepticism due to a lack of independent verification and benchmarks. While Alibaba suggested Qwen3.8 Max ranks second only to Anthropic's Claude Fable 5, this assertion was made in a single social media post without supporting data. Benchmark results from other sources suggest the model underperforms several domestic and international competitors. The Qwen3.8 Max-Preview is a 2.4-trillion-parameter sparse Mixture-of-Experts model with multimodal input support and a large context window of 983,616 tokens. The model's thinking capability is active by default with three intensity settings.
Alibaba's classification as a Chinese Military Company by the Pentagon in June 2026, which the company is contesting, adds another layer of scrutiny for potential enterprise adopters, alongside concerns about adherence to China's National Intelligence Law.
Pick the topics you care about. Get only what matters, on your cadence.