Pick the right Chinese AI model (GLM, Qwen, MiniMax, Seed, MiMo, DeepSeek): specs, PowerTokens pricing and a cost calculator.
Live site: https://powertokens.github.io/chinese-model-picker/
One sortable table compares context length, max output, tool calling, reasoning, input types and price per 1M tokens, with filters for what you need and a monthly cost estimator. Static HTML/CSS/JS, no build step, served by GitHub Pages.
python3 -m http.server 8765
# open http://localhost:8765/python3 scripts/fetch_prices.py # uses a local models.dev clone if MODELS_DEV_DIR / --models-dev is set,
python3 scripts/fetch_prices.py --models-dev ../models.dev # otherwise fetches lab files from GitHub rawWrites data/models.json. Python 3.11+, standard library only, no API keys. A weekly GitHub Action (.github/workflows/refresh-data.yml) runs the same script and commits data/models.json when models or prices change.
| Field | Source |
|---|---|
| Model list | Text models in docs.powertokens.ai/llms.txt that also have a published price |
| Price (USD / 1M tokens, incl. prompt-length tiers) | powertokens.ai/pricing (public endpoint POST /api/model/pricing) |
| Context, max output, tool calling, reasoning, input modalities, open weights, description | models.dev lab files models/<lab>/<model>.toml |
| "Best for" tags | Derived in scripts/fetch_prices.py from the fields above (see the Method section on the page) |
Specs are the lab's model specs; what a given API exposes can differ. Prices change; verify on the official pages.
Which Chinese AI model is best for coding?
No single model is best for every codebase. In this table, the models whose models.dev descriptions mention coding include Qwen3 Coder Plus, Qwen3 Max and the Qwen3.7 and Qwen3.8 Plus/Max models, GLM-4.5 Air, GLM-4.7 and the GLM-5.x series, MiniMax M2.5, M2.7 and M3, and Xiaomi MiMo-V2.5 and V2.5-Pro. Use the Coding filter to compare their context length, tool calling and price, then test the shortlist on your own code.
How do I use Qwen or GLM outside China?
Qwen is available internationally through Alibaba Cloud Model Studio and GLM through Z.ai; the mainland sites (Bailian, BigModel) may require a Chinese phone number, real-name verification or local payment. You can also call both through an OpenAI-compatible gateway such as PowerTokens, which serves every model in this table from https://api.powertokens.ai/v1.
Is there one API key for GLM, Qwen, MiniMax and DeepSeek V3.2?
Yes. A PowerTokens API key works for all models listed here, including GLM, Qwen, MiniMax, DeepSeek V3.2, ByteDance Seed and Xiaomi MiMo, and you switch models by changing the model string. On the official platforms each lab is a separate account, key and balance.
How much does it cost to use these models?
You pay per token, priced in USD per 1M input and output tokens, and some models charge a higher rate for long prompts. Enter your monthly token volume in the cost calculator to estimate the cost per model, and check powertokens.ai/pricing for current rates.
Are these models OpenAI-compatible?
Through PowerTokens, yes: every model here uses the OpenAI Chat Completions format, so the official OpenAI Python or Node SDK works with base_url set to https://api.powertokens.ai/v1. Click Use on any row to get a ready-to-run snippet.
Where do the specs and prices come from?
Context length, max output, tool calling, reasoning and input types come from the open models.dev catalog, and prices come from the PowerTokens pricing page. Both are regenerated into data/models.json by scripts/fetch_prices.py; the check date is shown in the footer.
scripts/screenshots.py (Playwright) writes review PNGs to screenshots/ (not committed).
MIT. Model names and trademarks belong to their owners.
Created by @uselesssoso for PowerTokens.