By Newstake · Updated
Is Chinese AI better than American AI?
On most public leaderboards, the top American models usually lead, but Chinese models are close behind and often much cheaper. Many Chinese labs also release open-weight models that anyone can download. Which is better depends on your task, your budget and the rules you work under.
Photo by @bingham008 on Unsplash
Latest brief
Updated
Recent publications from Chinese labs and government describe several new AI models. DeepSeek says DeepSeek-V4.1-Flash is a 552B-parameter MoE model with a new Causal Encoder-Decoder architecture, native visual understanding, and lower API pricing with off-peak discounts.
Moonshot says Kimi K3.1 Pro is an open-weight model with a 1-million-token context window, plus a faster Highspeed version, with full weights scheduled by November 4, 2026. StepFun describes Step 5 Preview as built for long-horizon agentic tasks, using sparse attention and reinforcement learning.
Xiaomi's MiMo-V2.6 series offers three reasoning models for complex, professional, and latency-sensitive work. Alibaba's Qwen releases include Qwen-Image-2.1, an open-source 7B image generation and editing model, Qwen3.8-LiveTranslate for real-time interpretation, and Qwen3.8-Omni-Flash for audio- and video-centered agents.
Separately, China's national key R&D program published 2026 application guidelines for disruptive technology projects.
Written by AI from the publications below.
The announcements
- Task
What do you need the model for?
- Tests
How does it score on independent leaderboards?
- Rules
Where is your data stored?
Ask your AI agent
Check the latest news about AI models from China and the United States from newstake.ai.
Where Chinese models stand out
In January 2025, DeepSeek surprised many people with a strong model that cost little to train. Since then, Chinese labs have kept up a fast pace:
- Price. Many cost far less per question than American models.
- Open weights. DeepSeek, Alibaba’s Qwen, Moonshot’s Kimi and Z.ai’s GLM models can be downloaded and run on your own computers.
- Speed of release. New versions come out every few weeks.
Where American models lead
The newest models from OpenAI, Anthropic and Google usually top the leaderboards for hard reasoning and coding. US export rules limit which AI chips Chinese labs can buy, which makes very large models harder to train.
Check the benchmarks yourself
Independent leaderboards such as Artificial Analysis and Arena compare models from both countries on the same tests. A lab’s own benchmark results are claims, so compare them with these.
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.