At the World Artificial Intelligence Conference in Shanghai, Alibaba introduced Qwen3.8, a new AI system designed to handle multimodal inputs like images, video, and text. The company asserts it ranks second only to Anthropic's Claude Fable 5 among global leading models but has yet to share benchmark data.
- Qwen3.8 features 2.4 trillion parameters and supports multimodal input
- Alibaba claims it is second only to Anthropic's Claude Fable 5
- No independent benchmarks or detailed performance data released yet
What happened
Alibaba Group previewed its newest AI model, Qwen3.8, at the World Artificial Intelligence Conference held in Shanghai. This release marks a significant jump in the Qwen family, as the model incorporates 2.4 trillion parameters, making it the first in Alibaba’s lineup to surpass the trillion-parameter threshold. It is designed to process various data types—images, videos, documents, and text—representing a major expansion in capability.
The Qwen3.8 model preview, named Qwen3.8-Max-Preview, is accessible through Alibaba’s subscription and developer platforms at a discounted trial price. The model is described as evolving continuously. However, Alibaba has not shared essential details such as benchmark scores, model cards, or the count of activated parameters. The company asserts that Qwen3.8 ranks just behind Anthropic’s Claude Fable 5 but provides no independent validation.
Why it matters
Alibaba's announcement situates it prominently in the fast-moving AI landscape, particularly within China, where competition is intensifying among developers pushing multi-trillion-parameter models. Reaching this scale aligns Alibaba with leading global AI labs, underscoring its ambition to become the default AI provider within the Chinese market and beyond. The open-weight strategy adopted by Alibaba has helped it build a sizable developer community around the Qwen family, promising broader adoption and innovation.
However, the notable absence of publicly available performance benchmarks or open weights for Qwen3.8 invites caution in validating Alibaba’s claim of ranking second to Claude Fable 5. Previous models in the Qwen series, like Qwen3.7-Max, were benchmarked and ranked significantly lower than Anthropic's top-tier offerings. Without transparent evaluations, enterprise buyers and the AI community must assess this new release on limited information.
What to watch next
Key indicators to monitor involve Alibaba releasing detailed benchmark results and publishing open weights for Qwen3.8, which the company has promised but not yet delivered. These will be crucial to verify the model’s performance and position relative to global leaders like Anthropic. Additionally, observing adoption trends among developers and enterprises subscribing to the Token Plan and Qoder platforms will signal market confidence.
Concurrently, the AI race in China is accelerating, with players like Beijing Moonshot AI Technology pushing even larger parameter models such as Kimi K3. How Alibaba’s offering stacks up against these domestic rivals and its standing on independent leaderboards will shape perceptions of China’s capability to compete at the top tier of AI innovation.