How to compare Chinese LLM API pricing

As of 2026-11-04.

Never mix CNY domestic list prices with USD international quotes. Prefer the console matching your billing entity.

Official pricing entry points

FamilyOfficial pricing URLCurrency note
DeepSeekhttps://api-docs.deepseek.com/quick_start/pricingUSD
Qwenhttps://www.alibabacloud.com/help/en/model-studio/model-pricingUSD (international Model Studio)
GLM (Zhipu / Z.AI)https://docs.z.ai/guides/overview/pricingUSD
Kimi (Moonshot)https://platform.moonshot.cn/docs/pricingCNY or USD depending on account region
MiniMaxhttps://www.minimax.io/pricingUSD
Doubao / Seed (ByteDance)https://www.byteplus.com/en/pricing?product=ModelArkUSD (BytePlus)
ERNIE (Baidu)https://intl.cloud.baidu.comDepends on region (CNY domestic / intl cloud billing)
Hunyuan (Tencent)https://www.tencentcloud.com/document/productDepends on Tencent Cloud region
StepFun (Step)https://platform.stepfun.ai/docs/en/guides/pricing/detailsUSD

Verified example: DeepSeek (transcribed)

From DeepSeek Models & Pricing as of 2026-11-04: deepseek-flash and deepseek-v4-pro bill per 1M tokens with cache hit/miss and peak/off-peak multipliers. Off-peak is half of peak. Peak windows: 01:00–04:00 and 06:00–10:00 UTC, Monday–Friday, excluding Chinese public holidays.

Verified example: StepFun (transcribed)

From StepFun pricing details: e.g. step-3.5-flash $0.10 / $0.02 / $0.30 (miss/hit/output per 1M); step-5-preview $1.00 / $0.10 / $2.70.

FAQ

Why does this page not list one unified price table?

Vendors use different units (cache hit/miss, peak/off-peak, per-minute audio, per-request search). A fake unified table would invent numbers. We link each official table and only transcribe figures we verified.