Chinese AI API Pricing Comparison 2026

Pay-as-you-go API pricing for 20+ Chinese AI models — DeepSeek, Kimi, GLM, Qwen, MiniMax and more. Compare input/output cost per million tokens, context windows, and capabilities.
📅 Continuously Updated · 2026🏷️ 10 Providers · 20 Models📌 Last update: July 4, 2026

Compare Chinese AI API Pricing 2026: Cost per Million Tokens

Building on Chinese AI models via API? We compared 20+ models from 10 providers side-by-side — including DeepSeek V4, Kimi K2.6, GLM-5.1, Qwen3.7, MiniMax M3, Doubao, Hunyuan, ERNIE and Spark. All prices are per million tokens (input / output), with context windows and capability tags. Updated July 2026.

20 models
ModelProviderInput /MOutput /MContextCapabilitiesNotes
DeepSeek V4 ProDeepSeek¥3.00Cache hit ¥0.025/M¥6.001M
TextFunction CallingStructured Output
CNYPeak hours (9-12 / 14-18) price doubles; GA in mid-July
DeepSeek V4 FlashDeepSeek¥1.00Cache hit ¥0.02/M¥2.001M
TextFunction CallingStructured Output
CNYPeak hours (9-12 / 14-18) price doubles
MiniMax M3MiniMax$0.30$1.201M
TextVisionVideoFunction Calling
USD
Kimi K2.6Kimi (Moonshot AI)¥6.50¥27.00262K
TextVisionVideoFunction Calling
CNY
Kimi K2.5Kimi (Moonshot AI)¥4.00¥21.00262K
TextVisionVideoFunction Calling
CNY
Kimi K2.7 CodeKimi (Moonshot AI)¥6.50¥27.00262K
TextFunction Calling
CNYThinking-only mode
GLM-5.1Zhipu AI (GLM)$1.40$4.40200K
TextFunction CallingStructured Output
USD
GLM-5 TurboZhipu AI (GLM)$1.20$4.00200K
TextFunction CallingStructured Output
USD
GLM-5V TurboZhipu AI (GLM)$1.20$4.00200K
TextVisionVideoFunction CallingStructured Output
USD
Qwen3.7 MaxAlibaba Cloud (Qwen)¥12.00¥36.001M
TextVisionFunction CallingStructured Output
CNY
Qwen3.7 PlusAlibaba Cloud (Qwen)¥2.00¥8.001M
TextVisionFunction CallingStructured Output
CNYContext >256K billed separately
MiMo-V2.5-ProXiaomi (MiMo)$0.435$0.871M
TextFunction CallingStructured Output
USD (≈¥3.00 / ≈¥6.00)MIT open-source
MiMo-V2.5Xiaomi (MiMo)$0.14$0.281M
TextVisionVideoAudioFunction Calling
USD (≈¥1.00 / ≈¥2.00)Omni-modal (text/image/video/audio)
Doubao-ProByteDance (Volcano Engine)¥1.00¥2.00128K
TextFunction Calling
CNY
Doubao-LiteByteDance (Volcano Engine)¥0.30¥0.60128K
TextFunction Calling
CNY
Hunyuan-LargeTencent Cloud (Hunyuan)¥4.00¥12.0032K
TextFunction Calling
CNY
Hunyuan-StandardTencent Cloud (Hunyuan)¥0.50¥1.0032K
TextFunction Calling
CNY
ERNIE 4.5Baidu (Qianfan)¥8.00¥24.0032K
TextFunction Calling
CNY
ERNIE SpeedBaidu (Qianfan)¥0.12¥0.128K
TextFunction Calling
CNY
Spark 4.0iFLYTEK (Spark)¥3.00¥9.00128K
TextFunction Calling
CNY
Capability legend:TextVisionVideoAudioFunction CallingStructured Output

📝 Pricing Notes

📌 Important Notes
• All prices are per million tokens (1M = 1,000,000 tokens) at official pay-as-you-go rates. Currency is shown per provider: ¥ = CNY (RMB), $ = USD.
• DeepSeek V4 Pro / V4 Flash implement peak/off-peak pricing: during peak hours (09:00–12:00 and 14:00–18:00, Beijing time) both input and output prices DOUBLE. Off-peak hours are half price — cross-timezone traffic can arbitrage this. DeepSeek also supports context caching: repeated prompt prefixes are billed at the much lower cache-hit rate (e.g. V4 Pro cache hit ¥0.025/M vs cache miss ¥3.00/M).
• Qwen3.7 Plus: input/output prices above apply to the first 256K context. Usage beyond 256K is billed separately at a higher long-context rate.
• Prices are subject to each platform's official announcements. Some providers offer additional volume discounts, tier packages, or free trial credits. Always verify before integration.