OpenClaw has been popular for a while. Also, because of network access and pricing, using OpenAI/Claude has a certain barrier and can be expensive. The rise of Chinese AI models, especially some open-source model releases, gives us more choices. This article looks at Chinese AI models and low-cost AI model options from several angles.
Recommended Chinese Models
Compared with the three major overseas models, GPT, Claude, and Gemini, there are also a few relatively good Chinese model providers. Personally, I am not considering much beyond these for now.
MiniMax
- I heard MiniMax is distilled from Claude. Based on my actual usage, it is definitely not as good as Claude/GPT, but compared with many alternatives, it is quite decent. The domestic pricing is especially cheap. For example, the coding plan starts at 29 RMB/month. If you upgrade the plan, you can also get image generation and audio capabilities, which I think is quite good.
🚀 MiniMax Token Plan is now available, adding voice, music, video, and image generation benefits. Invite friends to get double rewards and improve the development experience. Friends get an exclusive 10% discount + Builder benefits, while you get rebates + community privileges. The campaign only runs until the end of April, so use it while available. 👉 Join now: https://platform.minimaxi.com/subscribe/token-plan?code=56dZTYH77M&source=link
Kimi
- I am not using Kimi myself right now, but based on feedback from people around me, it is quite good, so it is also worth considering.
- Cursor also publicly acknowledged recently that Kimi is one of their base models. That is a kind of validation, so I think it is worth considering.
Zhipu GLM
- GLM is actually an established player. It raised prices recently, but usage is still increasing, which says something. Many people around me also use GLM, so it is worth considering.
I did not mention Qwen/Doubao and similar models here. I have used them, but personally I do not think they are as good as the options above, so I will not recommend them. Of course, they are still stronger than some models I mention below.
Cannot Decide? Use Them All
If you are unsure which Chinese AI model to choose, you can simply buy a cloud provider package. They usually provide access to all models. They are essentially resellers in some cases, or they deploy open-source models themselves, so there may be differences. But overall the gap is not large, because they are mostly using the same batch of models, and pricing is similar. The main question is which cloud provider’s service you prefer.
My Current Choice
I currently use the lowest tier of the MiniMax coding plan. If I need image generation later, I will upgrade to the next tier. Of course, MiniMax is only my fallback option. I mainly still use the overseas top three. If network access and payment are not problems, I still recommend using overseas models, because the gap is not just 5%.
Models I Personally Do Not Recommend
I have also tried some models that currently feel not good enough. I am listing them here so you can avoid wasting time. An unsuitable model not only wastes tokens and money, but also wastes a lot of life.
- Tencent Hunyuan
- DeepSeek
- DS is fine for translation, but do not expect much for coding. Maybe it will be much better when v4 comes out. Let’s wait and see.
Final Thoughts
- If you want to save time trying everything yourself, you can follow the recommendations above. If you want to try them all, that is also welcome.
