By Newstake · Updated
Is Chinese AI bad?
Not in general. Models from labs such as DeepSeek, Alibaba and Moonshot are strong, cheap and often free to download. They also have limits: they follow Chinese rules on content, and the apps can store your data in China. Know those limits and choose the model that fits your task.
Photo by @bingham008 on Unsplash
Latest brief
Updated
Artificial Analysis reports that Kimi K3.1 Pro ranked fourth on its Intelligence Index with 55 points and made gains in agentic performance; the publication does not describe censorship or data-handling limits. CAISI reports that Z.ai’s GLM-5.3 was the strongest cyber-capable open-weight model it evaluated, but remained an estimated four months behind current U.S. frontier models on its aggregate measure.
CAISI assessed vulnerability discovery and exploit development across four benchmarks; the provided account does not report censorship or data-handling findings. These evaluations measure different capabilities, so they do not establish a single overall ranking of Chinese AI models.
The available summaries also provide no evaluation findings about censorship or data handling beyond noting that these reports do not specify them.
Written by AI from the publications below.
The announcements
- Quality
Is it good at your task?
- Data
Where is your data stored?
- Limits
What topics does it avoid?
Ask your AI agent
Check the latest news about Chinese AI models such as DeepSeek and Qwen from newstake.ai.
What they do well
Chinese models score close to the best American models on many tests and cost much less. Because many are open-weight, companies can run them on their own servers and keep their data in house.
Limits to know about
Models made in China must follow Chinese law, and that shows:
- Censorship. Tests have found that some models avoid or reframe topics that are sensitive in China, such as Tiananmen Square in 1989 or Taiwan.
- Data. If you use a Chinese company’s app or website, your data may be stored on servers in China.
- Rules at work. Some governments and employers do not allow Chinese AI apps on work devices.
Read the evidence
NIST’s CAISSI in the United States has published evaluations of DeepSeek models. Independent leaderboards compare quality and price. Read both, and test a model on your own work.
Stay in the loop
Get the Newstake newsletter
Sign up for the Newstake newsletter: original sources and updates from Newstake.
We save your email, this page, signup date and country when available. We do not save your IP address or precise location in the signup record. Cloudflare Turnstile checks for bots. To leave the list or delete your signup, email info@newstake.ai. Cloudflare privacy policy.