计算公式
每月上游 Token 成本
请求数 × 天数 ×(非缓存输入 Token × 输入单价 + 缓存输入 Token × 缓存单价 + 输出 Token × 输出单价)÷ 1,000,000
建议月收入
(上游成本 + 平台成本)÷(1 − 目标净利率 − 支付费率 − 退款率 − 坏账率)
透明的数据与计算规则
PayAPIKey 公开 AI 模型价格参考所使用的来源、单位、核对日期和计算公式。计算器预设仅作为起点,不构成报价或账单。
仅当供应商第一方页面清晰列出模型标识和 Token 单价、足以进行统一换算时,才会收录该模型。
这些数据同时供 AI 模型成本计算器和公开 JSON 数据集使用。依赖某项价格前,请先访问其官方来源复核。
| 供应商 | 模型 / API ID | 标准价格 | 批处理价格 | 上下文 / 缓存规则 | 官方来源 |
|---|---|---|---|---|---|
| OpenAI | GPT-5.6 Solgpt-5.6-solga |
|
缓存写入: $3.125 | 缓存写入: $6.25 缓存存储: 未单独公布长上下文: 大于 272,000 输入 Token · 整次请求 Official standard and batch token rates; unpublished price components remain absent. Rates above 272k prompt tokens use the long-context tier.
缓存写入: $12.5 长上下文批处理
缓存写入: $6.25 | |
| OpenAI | GPT-5.6 Terragpt-5.6-terraga |
|
缓存写入: $1.25 | 缓存写入: $2.5 缓存存储: 未单独公布长上下文: 大于 272,000 输入 Token · 整次请求 Official standard and batch token rates; unpublished price components remain absent. Rates above 272k prompt tokens use the long-context tier.
缓存写入: $5 长上下文批处理
缓存写入: $2.5 | |
| OpenAI | GPT-5.6 Lunagpt-5.6-lunaga |
|
缓存写入: $0.125 | 缓存写入: $0.25 缓存存储: 未单独公布长上下文: 大于 272,000 输入 Token · 整次请求 Official standard and batch token rates; unpublished price components remain absent. Rates above 272k prompt tokens use the long-context tier.
缓存写入: $0.5 长上下文批处理
缓存写入: $0.25 | |
| OpenAI | GPT-5.5gpt-5.5ga |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: 未单独公布长上下文: 大于 272,000 输入 Token · 整次请求 Official standard and batch token rates; unpublished price components remain absent. Rates above 272k prompt tokens use the long-context tier.
缓存写入: 未单独公布 长上下文批处理
缓存写入: 未单独公布 | |
| OpenAI | GPT-5.5 Progpt-5.5-proga |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: 未单独公布长上下文: 大于 272,000 输入 Token · 整次请求 Official standard and batch token rates; unpublished price components remain absent. No long-context batch rate is published above 272k prompt tokens.
缓存写入: 未单独公布 长上下文批处理未单独公布 | |
| OpenAI | GPT-5.4gpt-5.4ga |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: 未单独公布长上下文: 大于 272,000 输入 Token · 整次请求 Official standard and batch token rates; unpublished price components remain absent. Rates above 272k prompt tokens use the long-context tier.
缓存写入: 未单独公布 长上下文批处理
缓存写入: 未单独公布 | |
| OpenAI | GPT-5.4 Minigpt-5.4-miniga |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: 未单独公布长上下文: 未单独公布Official standard and batch token rates; unpublished price components remain absent. | |
| OpenAI | GPT-5.4 Nanogpt-5.4-nanoga |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: 未单独公布长上下文: 未单独公布Official standard and batch token rates; unpublished price components remain absent. | |
| OpenAI | GPT-5.4 Progpt-5.4-proga |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: 未单独公布长上下文: 大于 272,000 输入 Token · 整次请求 Official standard and batch token rates; unpublished price components remain absent. Rates above 272k prompt tokens use the long-context tier.
缓存写入: 未单独公布 长上下文批处理
缓存写入: 未单独公布 | |
| OpenAI | GPT-5.2gpt-5.2legacy |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: 未单独公布长上下文: 未单独公布Official standard and batch token rates; unpublished price components remain absent. This served model is marked legacy, not retired. | |
| OpenAI | GPT-5.2 Progpt-5.2-prolegacy |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: 未单独公布长上下文: 未单独公布Official standard and batch token rates; unpublished price components remain absent. This served model is marked legacy, not retired. | |
| OpenAI | GPT-5.1gpt-5.1legacy |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: 未单独公布长上下文: 未单独公布Official standard and batch token rates; unpublished price components remain absent. This served model is marked legacy, not retired. | |
| OpenAI | GPT-4.1gpt-4.1legacy |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: 未单独公布长上下文: 未单独公布Official standard and batch token rates; unpublished price components remain absent. This served model is marked legacy, not retired. | |
| OpenAI | GPT-4.1 Minigpt-4.1-minilegacy |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: 未单独公布长上下文: 未单独公布Official standard and batch token rates; unpublished price components remain absent. This served model is marked legacy, not retired. | |
| OpenAI | GPT-4ogpt-4olegacy |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: 未单独公布长上下文: 未单独公布Official standard and batch token rates; unpublished price components remain absent. This served model is marked legacy, not retired. | |
| OpenAI | GPT-4o Minigpt-4o-minilegacy |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: 未单独公布长上下文: 未单独公布Official standard and batch token rates; unpublished price components remain absent. This served model is marked legacy, not retired. | |
| Anthropic | Claude Fable 5claude-fable-5ga |
|
缓存写入: 未单独公布 | 缓存写入(5 分钟): $12.5缓存写入(1 小时): $20 缓存存储: 未单独公布长上下文: 未单独公布Official token rates with distinct cache-read, five-minute write, and one-hour write prices. | |
| Anthropic | Claude Opus 5claude-opus-5ga |
|
缓存写入: 未单独公布 | 缓存写入(5 分钟): $6.25缓存写入(1 小时): $10 缓存存储: 未单独公布长上下文: 未单独公布Official token rates with distinct cache-read, five-minute write, and one-hour write prices. | |
| Anthropic | Claude Opus 4.8claude-opus-4-8ga |
|
缓存写入: 未单独公布 | 缓存写入(5 分钟): $6.25缓存写入(1 小时): $10 缓存存储: 未单独公布长上下文: 未单独公布Official token rates with distinct cache-read, five-minute write, and one-hour write prices. | |
| Anthropic | Claude Opus 4.7claude-opus-4-7ga |
|
缓存写入: 未单独公布 | 缓存写入(5 分钟): $6.25缓存写入(1 小时): $10 缓存存储: 未单独公布长上下文: 未单独公布Official token rates with distinct cache-read, five-minute write, and one-hour write prices. | |
| Anthropic | Claude Opus 4.6claude-opus-4-6ga |
|
缓存写入: 未单独公布 | 缓存写入(5 分钟): $6.25缓存写入(1 小时): $10 缓存存储: 未单独公布长上下文: 未单独公布Official token rates with distinct cache-read, five-minute write, and one-hour write prices. | |
| Anthropic | Claude Opus 4.5claude-opus-4-5-20251101ga |
|
缓存写入: 未单独公布 | 缓存写入(5 分钟): $6.25缓存写入(1 小时): $10 缓存存储: 未单独公布长上下文: 未单独公布Official token rates with distinct cache-read, five-minute write, and one-hour write prices. | |
| Anthropic | Claude Sonnet 5claude-sonnet-5ga |
|
缓存写入: 未单独公布 | 缓存写入(5 分钟): $2.5缓存写入(1 小时): $4 缓存存储: 未单独公布长上下文: 未单独公布优惠价有效至 2026-08-31下次价格生效日期 2026-09-01 Official token rates with distinct cache-read, five-minute write, and one-hour write prices. Introductory pricing ends after 2026-08-31.
缓存写入(5 分钟): $3.75缓存写入(1 小时): $6 批处理价格
缓存写入: 未单独公布 | |
| Anthropic | Claude Sonnet 4.6claude-sonnet-4-6ga |
|
缓存写入: 未单独公布 | 缓存写入(5 分钟): $3.75缓存写入(1 小时): $6 缓存存储: 未单独公布长上下文: 未单独公布Official token rates with distinct cache-read, five-minute write, and one-hour write prices. | |
| Anthropic | Claude Sonnet 4.5claude-sonnet-4-5-20250929ga |
|
缓存写入: 未单独公布 | 缓存写入(5 分钟): $3.75缓存写入(1 小时): $6 缓存存储: 未单独公布长上下文: 未单独公布Official token rates with distinct cache-read, five-minute write, and one-hour write prices. | |
| Anthropic | Claude Haiku 4.5claude-haiku-4-5-20251001ga |
|
缓存写入: 未单独公布 | 缓存写入(5 分钟): $1.25缓存写入(1 小时): $2 缓存存储: 未单独公布长上下文: 未单独公布Official token rates with distinct cache-read, five-minute write, and one-hour write prices. | |
Gemini 3.6 Flashgemini-3.6-flashga |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: $1 · USD per 1M cached tokens per hour长上下文: 未单独公布Official paid-tier text token rates; hourly context-cache storage is recorded separately. | ||
Gemini 3.5 Flashgemini-3.5-flashga |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: $1 · USD per 1M cached tokens per hour长上下文: 未单独公布Official paid-tier text token rates; hourly context-cache storage is recorded separately. | ||
Gemini 3.5 Flash-Litegemini-3.5-flash-litega |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: $1 · USD per 1M cached tokens per hour长上下文: 未单独公布Official paid-tier text token rates; hourly context-cache storage is recorded separately. | ||
Gemini 3.1 Flash-Litegemini-3.1-flash-litega |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: $1 · USD per 1M cached tokens per hour长上下文: 未单独公布Official paid-tier text token rates; hourly context-cache storage is recorded separately. | ||
Gemini 3.1 Pro Previewgemini-3.1-pro-previewpreview |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: $4.5 · USD per 1M cached tokens per hour长上下文: 大于 200,000 输入 Token · 整次请求 Official paid-tier text token rates; hourly context-cache storage is recorded separately. Rates above 200k prompt tokens use the long-context tier.
缓存写入: 未单独公布 长上下文批处理
缓存写入: 未单独公布 | ||
Gemini 3 Flash Previewgemini-3-flash-previewpreview |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: $1 · USD per 1M cached tokens per hour长上下文: 未单独公布Official paid-tier text token rates; hourly context-cache storage is recorded separately. | ||
Gemini 2.5 Progemini-2.5-proga |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: $4.5 · USD per 1M cached tokens per hour长上下文: 大于 200,000 输入 Token · 整次请求 计划停用日期: 2026-10-16Official paid-tier text token rates; hourly context-cache storage is recorded separately. Scheduled API shutdown is 2026-10-16.
缓存写入: 未单独公布 长上下文批处理
缓存写入: 未单独公布 | ||
Gemini 2.5 Flashgemini-2.5-flashga |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: $1 · USD per 1M cached tokens per hour长上下文: 未单独公布计划停用日期: 2026-10-16Official paid-tier text token rates; hourly context-cache storage is recorded separately. Scheduled API shutdown is 2026-10-16. | ||
Gemini 2.5 Flash-Litegemini-2.5-flash-litega |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: $1 · USD per 1M cached tokens per hour长上下文: 未单独公布计划停用日期: 2026-10-16Official paid-tier text token rates; hourly context-cache storage is recorded separately. Scheduled API shutdown is 2026-10-16. | ||
| xAI | Grok 4.5grok-4.5ga |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: 未单独公布长上下文: 大于或等于 200,000 输入 Token · 整次请求 Official text token rates; long-context pricing applies to the whole request at the threshold.
缓存写入: 未单独公布 长上下文批处理未单独公布 | |
| xAI | Grok Build 0.1grok-build-0.1preview |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: 未单独公布长上下文: 大于或等于 200,000 输入 Token · 整次请求 Official text token rates; long-context pricing applies to the whole request at the threshold.
缓存写入: 未单独公布 长上下文批处理未单独公布 | |
| xAI | Grok 4.3grok-4.3ga |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: 未单独公布长上下文: 大于或等于 200,000 输入 Token · 整次请求 Official text token rates; long-context pricing applies to the whole request at the threshold.
缓存写入: 未单独公布 长上下文批处理
缓存写入: 未单独公布 | |
| xAI | Grok 4.20 Reasoninggrok-4.20-0309-reasoningga |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: 未单独公布长上下文: 大于或等于 200,000 输入 Token · 整次请求 Official text token rates; long-context pricing applies to the whole request at the threshold.
缓存写入: 未单独公布 长上下文批处理
缓存写入: 未单独公布 | |
| xAI | Grok 4.20 Non-Reasoninggrok-4.20-0309-non-reasoningga |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: 未单独公布长上下文: 大于或等于 200,000 输入 Token · 整次请求 Official text token rates; long-context pricing applies to the whole request at the threshold.
缓存写入: 未单独公布 长上下文批处理
缓存写入: 未单独公布 | |
| xAI | Grok 4.20 Multi-Agentgrok-4.20-multi-agent-0309preview |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: 未单独公布长上下文: 大于或等于 200,000 输入 Token · 整次请求 Official text token rates; long-context pricing applies to the whole request at the threshold.
缓存写入: 未单独公布 长上下文批处理
缓存写入: 未单独公布 | |
| DeepSeek | DeepSeek V4 Flashdeepseek-v4-flashpreview |
| 未单独公布 | 缓存写入: 未单独公布 缓存存储: 未单独公布长上下文: 未单独公布价格通知: Peak-hour pricing at 2x has been announced without an effective date and is not active.Official cache-hit, cache-miss, and output rates; cache writes and storage are not independently priced. | |
| DeepSeek | DeepSeek V4 Prodeepseek-v4-propreview |
| 未单独公布 | 缓存写入: 未单独公布 缓存存储: 未单独公布长上下文: 未单独公布价格通知: Peak-hour pricing at 2x has been announced without an effective date and is not active.Official cache-hit, cache-miss, and output rates; cache writes and storage are not independently priced. | |
| Mistral | Mistral Medium 3.5mistral-medium-3-5ga |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: 未单独公布长上下文: 未单独公布Official text token rates; no combined batch-cache price is inferred from separate controls. | |
| Mistral | Mistral Small 4mistral-small-2603ga |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: 未单独公布长上下文: 未单独公布Official text token rates; no combined batch-cache price is inferred from separate controls. | |
| Mistral | Mistral Large 3mistral-large-2512ga |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: 未单独公布长上下文: 未单独公布Official text token rates; no combined batch-cache price is inferred from separate controls. | |
| Mistral | Ministral 3 3Bministral-3b-2512ga |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: 未单独公布长上下文: 未单独公布Official text token rates; no combined batch-cache price is inferred from separate controls. | |
| Mistral | Ministral 3 8Bministral-8b-2512ga |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: 未单独公布长上下文: 未单独公布Official text token rates; no combined batch-cache price is inferred from separate controls. | |
| Mistral | Ministral 3 14Bministral-14b-2512ga |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: 未单独公布长上下文: 未单独公布Official text token rates; no combined batch-cache price is inferred from separate controls. | |
| Mistral | Codestralcodestral-2508ga |
|
缓存写入: 未单独公布 | 缓存写入: 未单独公布 缓存存储: 未单独公布长上下文: 未单独公布Official text token rates; no combined batch-cache price is inferred from separate controls. |
USD / 1,000,000 tokens
请求数 × 天数 ×(非缓存输入 Token × 输入单价 + 缓存输入 Token × 缓存单价 + 输出 Token × 输出单价)÷ 1,000,000
(上游成本 + 平台成本)÷(1 − 目标净利率 − 支付费率 − 退款率 − 坏账率)
抓取程序、分析人员和潜在买家都可以核验界面所使用的同一组公开输入。
参考数据不会自动加入供应商的所有特殊收费。工具调用、搜索、图片、音频、视频、缓存写入、长上下文溢价、批处理折扣、地区处理、税费、汇率和合同价需要另行填写或核对。