LLM API Pricing

Per-million-token API prices for mainstream LLMs. Global (USD) and mainland-China (CNY) prices sit side by side per model, each with its official source and verification date.

💡 TL;DRCheapest output: Qwen-Turbo (≈$0.20/1M); priciest: Codex / GPT-5.5 API (≈$30/1M). Biggest cache discount: DeepSeek V4 Pro (~120× off).
Model🌍 Global USD🇨🇳 Mainland China CNYSource
InputHitOutputInputHitOutput
Qwen-Turbounverified
通义千问 Qwen· qwen.ai
输出价待核实
$0.05$0.2
MiMo V2.5
小米 MiMo· platform.xiaomimimo.com
¥1¥0.02¥2
DeepSeek V4 Flash
DeepSeek· deepseek.com
deepseek-chat=V4 Flash 非思考别名;全球统一价(约 ¥1/¥0.02/¥2)
$0.14$0.0028$0.28
GPT-4o mini
OpenAI· openai.com
$0.15$0.075$0.6
MiMo V2.5 Pro
小米 MiMo· platform.xiaomimimo.com
¥3¥0.025¥6
DeepSeek V4 Pro
DeepSeek· deepseek.com
deepseek-reasoner=V4 Pro 思考别名;全球统一价(约 ¥3/¥0.025/¥6)
$0.435$0.0036$0.87
Doubao Seed 1.6
字节 豆包· volcengine.com
0-32K 档;32K-128K ¥1.2/¥16;128K-256K ¥2.4/¥24。缓存价待核实
¥0.8¥8
Step 3.7 Flash
阶跃 StepFun· stepfun.com
¥1.35¥0.27¥8.1
MiniMax-M3
MiniMax· minimaxi.com
标准档 ≤512K 输入;>512K 为 $0.60/$0.12/$2.40
$0.3$0.06$1.2¥2.1¥0.42¥8.4
GPT-4.1 mini
OpenAI· openai.com
$0.4$0.1$1.6
Qwen-Plus
通义千问 Qwen· qwen.ai
$0.4$2.4
Grok 4.3
xAI· x.ai
grok-4/grok-3 等旧别名现路由至 Grok 4.3
$1.25$0.2$2.5
GLM-5.1
智谱 Zhipu· z.ai
输入 [0,32K) 档;[32K+) 为 ¥8/¥2/¥28
$1.4$4.4¥6¥1.3¥24
Kimi K2.6
Moonshot Kimi· kimi.com
¥6.5¥1.1¥27
Kimi K2.7 Code
Moonshot Kimi· kimi.com
262K 上下文;kimi-for-coding 即 K2.7 Code
$0.95$0.19$4¥6.5¥1.3¥27
Qwen3.7-Max
通义千问 Qwen· qwen.ai
Together/OpenRouter USD 路线;阿里官方为 RMB/分区价
$1.25$0.13$3.75
GLM-5.2
智谱 Zhipu· z.ai
1M 上下文;缓存存储限时免费
$1.4$0.26$4.4¥8¥2¥28
Claude Haiku 4.5
Anthropic· anthropic.com
$1$0.1$5
Kimi K2.7 Code highspeed
Moonshot Kimi· kimi.com
¥13¥2.6¥54
GPT-4.1
OpenAI· openai.com
$2$0.5$8
Kimi K2.7 Code Highspeed
Moonshot Kimi· kimi.com
$1.9$0.38$8
Gemini 3.5 Flash
Google· ai.google.dev
$1.5$0.15$9
Gemini 3.1 Pro Preview
Google· ai.google.dev
$2$0.2$12
Claude Sonnet 4.6
Anthropic· anthropic.com
$3$0.3$15
Claude Opus 4.8
Anthropic· anthropic.com
标准 API(不含 fast 模式);缓存命中约 9 折
$5$0.5$25
Claude Opus 4.6
Anthropic· anthropic.com
与 Opus 4.8 同价档;缓存命中约 9 折
$5$0.5$25
Codex / GPT-5.5 API
OpenAI· openai.com
标准短上下文 GPT-5.5 API(非 priority)
$5$0.5$30

Hit (cache-hit input) applies when context is reused (system prompts, long prefixes) and is usually far cheaper; “—” means the provider does not list it. Cross-currency comparisons use 1 USD ≈ 7.2 CNY. Prices change often — the official page prevails; “unverified” rows await confirmation.

Speed vs costMeasure cost live