Alibaba's Tiny New AI Models Punch Way Above Their Weight

Alibaba drops four open-weight small models, claims the 9B version goes toe-to-toe with OpenAI's 120B parameter model.

Alibaba's Tiny New AI Models Punch Way Above Their Weight

Alibaba just dropped the Qwen3.5 Small Model Series, and the numbers are wild. The lineup comes in four sizes — 0.8B, 2B, 4B, and 9B parameters — all released as open-weight models anyone can download and run.

The headline claim: that 9B model allegedly rivals OpenAI's gpt-oss-120b on certain benchmarks. That's a model roughly 13 times its size. If the benchmarks hold up under scrutiny, it's a significant efficiency win.

The release comes from Alibaba's dedicated Qwen Team, the e-commerce giant's AI research group that has been steadily shipping open models. Small, efficient models matter because they can run on consumer hardware and edge devices — no massive data center required.

Four models, all open-weight, with the smallest clocking in at just 800 million parameters. The AI efficiency race just got more interesting.