I've been experimenting with Chinese LLMs recently and the pricing caught me off guard:
| Model | Input (per 1M tokens) | Output (per 1M tokens) |
|---|---|---|
| GPT-4o | $2.50 | $10.00 |
| DeepSeek V4 | $0.30 | $0.50 |
| Qwen3.5-Plus | $0.11 | $0.67 |
| Qwen-Long | $0.07 | $0.28 |
That's 8-20x cheaper than GPT-4o depending on the model.
What's the Catch?
From my experience, the main barriers for non-Chinese developers are:
- Payment — Most providers only accept Alipay/WeChat Pay
- Documentation — English docs are limited or missing
- API compatibility — Each provider has its own format
- Trust — Concerns about data handling and compliance
Quality-wise?
I've been running side-by-side tests. For general tasks (chat, summarization, code assistance), DeepSeek and Qwen hold up surprisingly well. They're not beating GPT-4o on everything, but for the price? Hard to ignore.
I'm Curious
For those of you who have tried Chinese LLMs — what was your experience? And for those who haven't, what's holding you back?
Would love to hear different perspectives in the comments.