Live data, auto-updated daily · 2026-10-09
Choosing between Chinese AI models used to mean stitching together scattered pricing pages. We now publish a single live price-performance leaderboard covering 100+ Chinese frontier models behind one OpenAI-compatible endpoint, ranked cheapest-first within each category. This post explains what it measures and highlights the current leaders.
Every row is generated automatically from the live per-model pages on aiapi-pro.com (their Product structured data), refreshed daily. Prices are the raw upstream provider list rate passed through with 0% platform markup — unlike marketplaces that add ~5.5%. A machine-readable feed is available at /china-ai-leaderboard.json for tools, comparison sites and agents to consume directly.
| Category | Cheapest model | Price (USD) | Billing |
|---|---|---|---|
| Chat / LLM | GLM-4.7-Flash | Free | Free |
| Video | CogVideoX-Flash | Free | Free |
| Image | GLM-4.1V-Thinking-Flash | Free | Free |
| Audio | Qwen TTS | $0.013 | flat |
| 3D | Hunyuan 3D Motion | $0.185 | per generation |
| Embedding | Kinfra VL Embedding 2B | $0.0769 | per image |
These models cost nothing and have no token cap on NovAI — ideal for prototyping: GLM-4.7-Flash, GLM-4.6V-Flash, GLM-4.1V-Thinking-Flash, CogView3-Flash, CogVideoX-Flash. New accounts also receive $2 free credit with no card required.
Aggregators that add a markup make your cost unpredictable as usage scales. NovAI charges the provider rate exactly, so the leaderboard price is the price you pay. Video is billed per finished second or per upstream token; failed generations refund automatically.
Related: Cheapest Chinese LLM APIs in 2026 · NovAI vs OpenRouter · Leaderboard JSON feed