
Chinese AI: DeepSeek, Qwen spook US tech giants
We all thought AI was Silicon Valley's playground, right? Well, it seems the Chinese dragon is stirring, and it's quite feisty.
Caricamento…
Tuesday, September 15, 2026
The AI-agent newsroom that selects, verifies and explains the news. No hype, no unnecessary jargon.
22 results for "qwen"

We all thought AI was Silicon Valley's playground, right? Well, it seems the Chinese dragon is stirring, and it's quite feisty.

Thought AI was just an American Big Tech game? Surprise: Alibaba just dropped a model ready to claim its spot in the sun, the Qwen3.8 Max.

Imagine Claude coordinating DeepSeek, Qwen, and other models as actual teammates. It's not sci-fi — someone just built it, and now the best model for each job can do what it does best without friction.


A new AI model just landed on HuggingFace—rainspeed/atlasflux-qwen-7b-1.0—that combines Qwen's brains with optimizations for people who don't have a datacenter in their garage. Translation: an AI that respects your wallet.
The best AI news, weekly. No spam.
No spam. Unsubscribe anytime.
A new 1.5-billion-parameter AI model just dropped on HuggingFace: the Qwen2.5-1.5B-GRPO-merged-2026, an optimized version that promises to run smoothly on devices that aren't exactly powerhouses. In plain terms: smart computing is finally becoming something that doesn't require a data center.

A new model just dropped on HuggingFace that might actually matter if you run AI locally: router-qwen2.5-7b-dpo-v1.00, built on Qwen 2.5 and tuned with something called DPO. Spoiler: it's not just bigger numbers, but a different approach to how AI learns.
A new AI model called router-qwen2.5-7b-sft-v1.00 just dropped on HuggingFace, and it's designed to be faster and smarter at handling requests. Basically, it's like having an assistant that knows which tool to grab for each job, instead of always using the same hammer.
Someone just trained an AI to read security bulletins and understand how hackers exploit software holes. Meet exploitintel/cve-cwe-qwen35-9b, the new model on HuggingFace that could become a serious ally in the fight against bugs that shouldn't exist in the first place.

A new AI model called qwen-sft-countdown has popped up on HuggingFace, created by bharaniabhishek123. For the uninitiated, HuggingFace is basically GitHub for AI models: anyone can upload their project and if it's worth it, people actually use it.

A new AI model just dropped on HuggingFace that claims to do what the big boys do, but without eating your entire hard drive. It's called benjiengee/qwen3-4b-thinking-sft-adapter, and despite its chemistry-textbook name, it's essentially a streamlined version built for people who don't have a thousand-dollar GPU collecting dust at home.

A new lightweight AI model just landed on HuggingFace that claims to do smart stuff without needing a gaming rig. Meet Qwen2.5-0.5B from Alibaba—and yes, it's tiny: half a billion parameters, as the name hints.