US vs. China · tracked daily
US vs. China: real AI model pricing
Based on real per-token pricing, the median output-token price among tracked US flagship models is $4.00/1M, versus $1.20/1M for tracked Chinese flagship models - roughly 3.3x higher for US models. One flagship model per lab, picked by real Artificial Analysis quality score, not curated by hand.
Input and output token prices, per flagship model from each classified lab - the flagship is the model with the highest published Artificial Analysis Intelligence Index for that lab, not a manual pick. Origin is a curated allowlist by company headquarters (see server/origin-class.ts); labs with an ambiguous or non-US/non-China HQ (e.g. Mistral, Cohere) are excluded rather than guessed into either bucket.
| # | Model | Origin | Quality | Input/1M | Output/1M |
|---|---|---|---|---|---|
| 01 | Anthropic: Claude Opus 5.5 (batch) | US | 57.6 | $4.00 | $20.00 |
| 02 | Qwen: Qwen3.8 Max | China | 53.4 | $2.00 | $6.00 |
| 03 | OpenAI: GPT-5.4 (batch) | US | 53.1 | $2.50 | $15.00 |
| 04 | SpaceXAI: Grok 4.7 | US | 46.4 | $1.60 | $4.80 |
| 05 | Xiaomi: MiMo-V2.6-Pro | China | 46.3 | $0.43 | $0.87 |
| 06 | Z.ai: GLM 5.3 (batch) | China | 44.8 | $0.84 | $2.64 |
| 07 | MoonshotAI: Kimi K3 (batch) | China | 43.6 | $3.00 | $15.00 |
| 08 | Google: Gemini 3.8 Flash (batch) | US | 40.9 | $0.75 | $3.75 |
| 09 | Meta: Muse Spark 1.2 | US | 39.6 | $1.25 | $4.25 |
| 10 | DeepSeek: DeepSeek V4.1 Flash (batch) | China | 39.5 | $0.10 | $0.50 |
| 11 | Kwaipilot: KAT-Coder-Pro V2 | China | 33.7 | $0.30 | $1.20 |
| 12 | StepFun: Step 3.7 Flash | China | 30.9 | $0.20 | $1.15 |
| 13 | MiniMax: MiniMax M3 | China | 29.2 | $0.30 | $1.20 |
| 14 | Thinking Machines: Inkling Small (free) | US | 26.1 | $0.45 | $1.20 |
| 15 | Tencent: Hy3 preview | China | 25.3 | $0.18 | $0.60 |
| 16 | inclusionAI: Ling 3.0 Flash VL (free) | China | 24.6 | $0.06 | $0.18 |
| 17 | NVIDIA: Nemotron 3 Ultra (free) | US | 22.9 | $0.60 | $2.40 |
| 18 | Meituan: LongCat 2.0 | China | 19.1 | $0.30 | $1.20 |
| 19 | Amazon: Nova 2 Lite | US | 18.4 | $0.30 | $2.50 |
One model per lab keeps this a lab-vs-lab comparison, not a raw model count - a lab with five cheap small models wouldn't otherwise be comparable to a lab with one flagship. Quality is Artificial Analysis's published Intelligence Index; it is not adjusted or reweighted here.
Frequently asked questions
Are Chinese AI models cheaper than US AI models?
Yes, based on real pricing: the median output-token price among tracked US flagship models is $4.00/1M, versus $1.20/1M for tracked Chinese flagship models - about 3.3x higher for US models.
What is the cheapest US AI model and cheapest Chinese AI model?
Among the flagship models tracked here, Thinking Machines: Inkling Small (free) is the cheapest US model at $1.20/1M output tokens, and inclusionAI: Ling 3.0 Flash VL (free) is the cheapest Chinese model at $0.18/1M.