LLM API Pricing Table
Input / output / cache-hit prices (per 1M tokens) across models, with currency and tier tags — searchable and sortable by price.
| Model | Input /1M | Output /1M | Cache /1M | Source |
|---|---|---|---|---|
| GPT-5.5 OpenAI · USD | $5 | $30 | $0.5 | official |
| GPT-5.5 Batch OpenAI · USD | $2.5 | $15 | $0.25 | official |
| GPT-5.4 OpenAI · USD | $2.5 | $15 | $0.25 | official |
| GPT-5.4 mini OpenAI · USD | $0.75 | $4.5 | $0.075 | official |
| Claude Opus 4.8 Anthropic · USD | $5 | $25 | $0.5 write $6.25 | official |
| Claude Sonnet 4.6 Anthropic · USD | $3 | $15 | $0.3 write $3.75 | official |
| Claude Sonnet 4.6 Batch Anthropic · USD | $1.5 | $7.5 | $0.15 | official |
| Claude Haiku 4.5 Anthropic · USD | $1 | $5 | $0.1 write $1.25 | official |
| Claude Sonnet 4.5 Anthropic · USD | $3 | $15 | $0.3 write $3.75 | official |
| Gemini 3.5 Flash Google · USD | $1.5 | $9 | $0.15 | official |
| Gemini 3.5 Flash Batch Google · USD | $0.75 | $4.5 | $0.075 | official |
| Gemini 3.1 Pro Preview tiered Google · USD | <= 200,000 input$2> previous input$4 | <= 200,000 input$12> previous input$18 | <= 200,000 input$0.2> previous input$0.4 | official |
| Gemini 3.1 Pro Preview Batch tiered Google · USD | <= 200,000 input$1> previous input$2 | <= 200,000 input$6> previous input$9 | <= 200,000 input$0.2> previous input$0.4 | official |
| Gemini 3.1 Flash-Lite Google · USD | $0.25 | $1.5 | $0.025 | official |
| DeepSeek V4 Pro 官方人民币价 DeepSeek · CNY 中文官方价格页，按充值余额人民币计价。 | ¥3 | ¥6 | ¥0.025 | official |
| DeepSeek V4 Pro Official USD price DeepSeek · USD English official price page. | $0.435 | $0.87 | $0.0036 | official |
| DeepSeek V4 Flash 官方人民币价 DeepSeek · CNY 中文官方价格页，按充值余额人民币计价。 | ¥1 | ¥2 | ¥0.02 | official |
| DeepSeek V4 Flash Official USD price DeepSeek · USD English official price page. | $0.14 | $0.28 | $0.0028 | official |
| MiniMax-M3 Standard tiered MiniMax · USD Permanent 50% off pricing shown on official page; >512k availability may be limited. | <= 512,000 input$0.3> previous input$0.6 | <= 512,000 input$1.2> previous input$2.4 | <= 512,000 input$0.06> previous input$0.12 | official |
| MiniMax-M3 Priority tiered MiniMax · USD Priority tier: set service_tier=priority; official page says 1.5x standard. | <= 512,000 input$0.45> previous input$0.9 | <= 512,000 input$1.8> previous input$3.6 | <= 512,000 input$0.09> previous input$0.18 | official |
| MiniMax-M2.7 MiniMax · USD | $0.3 | $1.2 | $0.06 write $0.375 | official |
No matching model.
Pending official price checks
These models have official pages, but stable public input/output token prices were not visible in the fetched page content, so they are not used by the calculators yet.
- Kimi / Moonshot AI kimi-k2.7-code, kimi-k2.6 官方文档已确认模型页、计费单位和 Batch 60% 规则，但当前公开 HTML 未暴露具体输入/输出数字；暂不加入计算器，等能核验官方数字后再加。 official
- Alibaba Cloud Model Studio / Qwen qwen3.7-max, qwen3.7-plus, qwen3.6-flash 官方帮助页已确认模型存在，但价格明细在控制台动态页加载；暂不加入计算器，等能核验公开官方数字后再加。 official
- Xiaomi MiMo MiMo-V2.5-Pro, MiMo-V2.5, MiMo Code 小米 MiMo 官网已确认模型和 API 接入入口，但公开页面未展示 token 输入/输出价格；暂不加入计算器，等官方价格数字可核验后再加。 official
- Zhipu / GLM GLM-5.1, GLM-5, GLM-4.7 智谱开放平台价格页为动态页面，当前抓取结果未暴露具体 token 单价；暂不加入计算器，等官方价格数字可核验后再加。 official
A side-by-side LLM API pricing table and AI model price comparison: compare OpenAI, Claude, Gemini, DeepSeek and other model input, output, cached-input and tiered prices per 1M tokens, with billing currency, checked date and official source links. Pair it with the two cost calculators to compute actual spend.
Related search intents: LLM API pricing table · AI model pricing comparison · GPT pricing table · Claude API pricing · Gemini API pricing · DeepSeek API pricing · input output token price · cached input pricing · LLM token price per 1M · AI API price comparison · model pricing table · API token pricing · tiered model pricing · LLM cost comparison · OpenAI Anthropic Gemini pricing
FAQ
What unit are prices in?
All prices are per 1M tokens in the model’s billing currency — USD for USD models, CNY for CNY models. Use a calculator’s rate field to convert.
What does “tiered” mean?
The model switches price tier by per-call input or output length. The table shows all tiers; sorting uses the first tier, and the calculator selects the tier from your usage.
Why are some cache prices “—”?
It means the model has no public or modeled cache-hit price; it is billed at the input price and the hit rate has no effect.
How often are prices updated?
The table shows a checked date under the prices. This is not a provider billing system; if a provider just changed prices, its official pricing page wins.