anyone using china model? which one and any advise?

Posted by thisisqq@reddit | LocalLLaMA | View on Reddit | 9 comments

qwen from alibaba
ERNIE Bot from baidu
kimi from moonshot
deepseek
Doubao from byte dance

-----------

Actually, until recently, I've been using Anthropic's Claude with a VPN within China, but lately it's been getting blocked more and more often. So I'm reluctantly starting to consider Chinese AI models.

As for my usage, I don't do much actual coding — most of my work involves script writing, project structuring, market reserach, and business model simulations.

[-]

Immediate-Word1958@reddit

Also based in China. For your use case (script writing, project structuring, market research), I'd rank them:

DeepSeek — best overall for the price. V3.2 handles English really well, and the reasoning mode is solid for business analysis. API is dirt cheap ($0.28/M input tokens).
Qwen — very close second, especially Qwen3.5 Plus. Slightly better at structured outputs and following complex instructions in my experience. Alibaba Cloud (Bailian) platform is a bit clunky but works.
Kimi — good for long document analysis (200K context). Worth trying for market research with lots of source material.
Doubao — decent but nothing stands out vs DeepSeek/Qwen. The Volcengine platform is more enterprise-oriented.
ERNIE — honestly I'd skip it. Feels a generation behind the others now.

For your workflow, I'd start with DeepSeek's web app (chat.deepseek.com) since it's free and you can test immediately. If you like it, the API is worth setting up.

[-]

Psyko38@reddit

I use Qwen on the cloud and locally. For complex reasoning tasks, Qwen 3.6 Plus is good on their UI, all unlimited without a subscription. For my part, I use it 50% with ChatGPT and locally for privacy-critical tasks.

[-]