clauding.de
DE
• 2 min read AI-generated

Qwen3.8: Alibaba benchmarks itself against Claude Fable 5 — with zero benchmarks

Copy article as Markdown

At WAIC in Shanghai, Alibaba previewed Qwen3.8-Max-Preview: 2.4 trillion parameters, multimodal for the first time — and a claim that only Anthropic's Claude Fable 5 beats it. The proof? None. No model card, no benchmarks, open weights 'soon'.

Alibaba Qwen Fable 5
Featured image for "Qwen3.8: Alibaba benchmarks itself against Claude Fable 5 — with zero benchmarks"

There’s one sentence from this week that tells you more about the AI race than any benchmark table. At the World Artificial Intelligence Conference in Shanghai, Alibaba unveiled its most powerful model yet — Qwen3.8 — and said it ranks “second only to Fable 5.” That’s Anthropic’s model. Not GPT-5.6, not Gemini. Claude.

What Alibaba is showing

The preview build is called Qwen3.8-Max-Preview and carries 2.4 trillion parameters. It’s the first Qwen model above the trillion mark that handles not just text but images, video and documents too. It’s available now through Alibaba’s Token Plan subscription and its Qoder and QoderWork developer platforms — at 10% of the standard price during the trial.

Sounds like a statement. It is one. It just comes without evidence.

The striking part: nothing

Alibaba’s Qwen team made the Fable 5 comparison in a post on X — without a single benchmark score. No model card. No activated-parameter count. No independent evaluation. That’s a sharp contrast to its predecessor: Qwen3.7-Max shipped in May with a full set of published results, including a 56.6 on the Artificial Analysis Intelligence Index.

This time? A model described only as “continuously evolving.” Open weights are promised “soon” — no date, no license terms.

The timing isn’t random. Chinese rival Moonshot released Kimi K3 with 2.8 trillion parameters three days earlier. Both launches push Chinese developers into the multi-trillion-parameter tier — a scale until recently reserved for the largest U.S. labs.

My take

For me, the most interesting news isn’t the model — it’s the reference point. When a Chinese giant unveils its flagship and picks Claude Fable 5 of all things as the bar to clear, that says something about how the market sees the field right now. Fable 5 is the number to beat — not GPT, not Gemini.

But a claim without numbers is a marketing statement, not a technical one. The last Qwen model that actually got ranked sits well down LMArena’s list — while Fable 5 is No. 1 there. Until Alibaba delivers benchmarks and those promised open weights, “second-best model in the world” stays a footnote, not a fact.

I still find these moments instructive. They show you where the race is being measured right now — and against whom. And they’re a reminder to keep a healthy dose of skepticism handy: when someone shows nothing, they often have nothing to show yet.

Sources: