Alibaba's Qwen team has previewed a new flagship artificial intelligence model, Qwen3.8-Max-Preview, and is making a bold claim about where it ranks among the world's most capable systems.
According to MarkTechPost, the model is a 2.4 trillion-parameter multimodal "mixture-of-experts" system — an architecture that routes each request to a subset of the model's parameters rather than firing all of them at once, which helps large models run more efficiently. "Multimodal" means it is designed to handle more than just text.
The headline claim is that Qwen3.8-Max is "second only to Fable 5," a reference to Anthropic's Claude Fable 5. SiliconANGLE summarizes the pitch the same way: Alibaba is positioning its model as bettered only by Claude.
For now, that is a claim without much to back it up. MarkTechPost notes that what is live is the preview itself — available on Alibaba's Token Plan, Qoder, and QoderWork services at 10% of standard pricing. What is not live, per MarkTechPost, is any benchmark table, model card, license, or per-token price. In other words, there is no published data yet for outsiders to verify the ranking.
The timing is notable. MarkTechPost points out the preview lands just days after rival Chinese AI lab Moonshot released its Kimi K3 model with open weights, underscoring how quickly China's leading labs are trading blows at the top of the field.
Why it matters: bold performance claims are cheap until benchmarks, pricing, and licenses arrive — so the real test of whether Alibaba can credibly rival Anthropic's Claude is the data it has not yet released.