AI API Token Price Calculator: Claude, GPT, Gemini, DeepSeek and More, per Thousand to per 100 Million Tokens
Pick a model and enter token counts to get the cost at each provider's official list price, converted between USD and CNY, or rank every model for the same usage. The table below lists input, cache-hit and output prices with a link to each official source, last checked on 2026-11-12. The formula is simple: cost = input tokens × input price + cached tokens × cache price + output tokens × output price, with prices per million tokens.
DeepSeek list price per million tokens: input ¥2 · cache hit ¥0.04 · output ¥8
Defaults to the CNY central parity rate on 2026-11-09. Change it to the rate you actually pay.
Cost of this usage
¥3.6
about $0.5358
Breakdown: Input (cache miss) ¥2 + Cache hit (cache read) ¥0 + Output ¥1.6
Per 1,000 tokens
input ¥0.002 · output ¥0.008
Per 100 million tokens
input ¥200 · output ¥800
Don't know the token count? Estimate from text length
Uses the rough ratios in DeepSeek's official docs: 1 Chinese character ≈ 0.6 tokens, 1 English character ≈ 0.3 tokens. Every model tokenizes differently; the usage field in the API response is the real count.
The same usage on every other model
Uses the token counts and exchange rate above and sorts every model from cheapest to most expensive.
| Model | Provider | List currency | In CNY | In USD |
|---|---|---|---|---|
| GLM-4.7-Flashglm-4.7-flash | Zhipu (GLM) | ¥0 | ¥0 | $0 |
| Qwen-Flashqwen-flash | Alibaba Cloud Model Studio (Qwen) | ¥0.45 | ¥0.45 | $0.067 |
| Doubao Seed 2.0 Minidoubao-seed-2-0-mini-260428 | Volcengine Ark (Doubao) | ¥0.6 | ¥0.6 | $0.0893 |
| Qwen-Plusqwen-plus | Alibaba Cloud Model Studio (Qwen) | ¥1.2 | ¥1.2 | $0.1786 |
| Qwen3.8-Flashqwen3.8-flash | Alibaba Cloud Model Studio (Qwen) | ¥1.34 | ¥1.34 | $0.1994 |
| Doubao Seed 2.1 Litedoubao-seed-2-1-lite-260915 | Volcengine Ark (Doubao) | ¥1.34 | ¥1.34 | $0.1994 |
| Claude Haiku 5.5claude-haiku-5-5 | Anthropic (Claude) | $0.2 | ¥1.34 | $0.2 |
| GPT-6 Lunagpt-6-luna | OpenAI (GPT) | $0.2 | ¥1.34 | $0.2 |
| GLM-5.3-Flashglm-5.3-flash | Zhipu (GLM) | ¥1.36 | ¥1.36 | $0.2024 |
| Qwen3-Coder-Flashqwen3-coder-flash | Alibaba Cloud Model Studio (Qwen) | ¥1.8 | ¥1.8 | $0.2679 |
| MiniMax-M3.1-FlashMiniMax-M3.1-Flash | MiniMax | ¥1.8 | ¥1.8 | $0.2679 |
| GPT-5.4 nanogpt-5.4-nano | OpenAI (GPT) | $0.45 | ¥3.02 | $0.45 |
| GLM-5.3-FlashXglm-5.3-flashx | Zhipu (GLM) | ¥3.4 | ¥3.4 | $0.506 |
| DeepSeek V4.1 Flashdeepseek-flash | DeepSeek | ¥3.6 | ¥3.6 | $0.5358 |
| Qwen3.7-Plusqwen3.7-plus | Alibaba Cloud Model Studio (Qwen) | ¥3.6 | ¥3.6 | $0.5358 |
| MiniMax-M3MiniMax-M3 | MiniMax | ¥3.78 | ¥3.78 | $0.5626 |
| Gemini 3.5 Flash-Litegemini-3.5-flash-lite | Google (Gemini) | $0.8 | ¥5.38 | $0.8 |
| Doubao Seed 2.1 Turbodoubao-seed-2-1-turbo-260628 | Volcengine Ark (Doubao) | ¥6 | ¥6 | $0.893 |
| Doubao Seed 2.0 Codedoubao-seed-2-0-code-preview-260215 | Volcengine Ark (Doubao) | ¥6.4 | ¥6.4 | $0.9526 |
| GLM-5glm-5 | Zhipu (GLM) | ¥7.6 | ¥7.6 | $1.13 |
| Gemini 3.8 Flashgemini-3.8-flash | Google (Gemini) | $1.5 | ¥10.08 | $1.5 |
| GLM-5.1glm-5.1 | Zhipu (GLM) | ¥10.8 | ¥10.8 | $1.61 |
| GPT-5.4 minigpt-5.4-mini | OpenAI (GPT) | $1.65 | ¥11.09 | $1.65 |
| Kimi K2.6kimi-k2.6 | Moonshot AI (Kimi) | ¥11.9 | ¥11.9 | $1.77 |
| Doubao Seed 2.1 Prodoubao-seed-2-1-pro-260915 | Volcengine Ark (Doubao) | ¥12 | ¥12 | $1.79 |
| Claude Haiku 4.5claude-haiku-4-5-20251001 | Anthropic (Claude) | $2 | ¥13.44 | $2 |
| GPT-6.1 Terragpt-6.1-terra | OpenAI (GPT) | $2 | ¥13.44 | $2 |
| GLM-5.3glm-5.3 | Zhipu (GLM) | ¥13.6 | ¥13.6 | $2.02 |
| DeepSeek V4 Prodeepseek-v4-pro | DeepSeek | ¥14.4 | ¥14.4 | $2.14 |
| Qwen3.8-Maxqwen3.8-max | Alibaba Cloud Model Studio (Qwen) | ¥19.2 | ¥19.2 | $2.86 |
| Kimi K3.1 Prokimi-k3.1-pro | Moonshot AI (Kimi) | ¥26.6 | ¥26.6 | $3.96 |
| Claude Sonnet 5.5claude-sonnet-5-5 | Anthropic (Claude) | $4 | ¥26.88 | $4 |
| Claude Sonnet 5claude-sonnet-5 | Anthropic (Claude) | $4 | ¥26.88 | $4 |
| GPT-6.1 Solgpt-6.1-sol | OpenAI (GPT) | $4 | ¥26.88 | $4 |
| Gemini 4 Carbongemini-4-carbon | Google (Gemini) | $4 | ¥26.88 | $4 |
| Gemini 4 Argongemini-4-argon | Google (Gemini) | $4 | ¥26.88 | $4 |
| Gemini 3.1 Pro Previewgemini-3.1-pro-preview | Google (Gemini) | $4.4 | ¥29.56 | $4.4 |
| GPT-5.3 Codexgpt-5.3-codex | OpenAI (GPT) | $4.55 | ¥30.57 | $4.55 |
| GPT-5.4gpt-5.4 | OpenAI (GPT) | $5.5 | ¥36.95 | $5.5 |
| Kimi K3kimi-k3 | Moonshot AI (Kimi) | ¥40 | ¥40 | $5.95 |
| Claude Sonnet 4.6claude-sonnet-4-6 | Anthropic (Claude) | $6 | ¥40.31 | $6 |
| Claude Opus 5.5claude-opus-5-5 | Anthropic (Claude) | $8 | ¥53.75 | $8 |
| Claude Opus 5claude-opus-5 | Anthropic (Claude) | $10 | ¥67.19 | $10 |
| GPT-5.5gpt-5.5 | OpenAI (GPT) | $11 | ¥73.91 | $11 |
| Claude Fable 5.2claude-fable-5-2 | Anthropic (Claude) | $16 | ¥107.5 | $16 |
| Claude Fable 5.1claude-fable-5-1 | Anthropic (Claude) | $20 | ¥134.38 | $20 |
| GPT-6 Astragpt-6-astra | OpenAI (GPT) | $20 | ¥134.38 | $20 |
Uses each provider's standard online list price (not batch or off-peak discounts). Models priced by input-length tier use the lowest tier; see the notes in the price table for long contexts. Your provider's bill is the final word.
Cost per 1,000, per million and per 100 million tokens
Each cell shows input / output in the provider's list currency, without cache discounts. To answer “how much are 1,000 tokens” or “how much are 100 million tokens”, read the matching column.
| Model | Provider | Per 1,000 tokens | Per million tokens | Per 100 million tokens |
|---|---|---|---|---|
| DeepSeek V4.1 Flashdeepseek-flash | DeepSeek | ¥0.002 / ¥0.008 | ¥2 / ¥8 | ¥200 / ¥800 |
| DeepSeek V4 Prodeepseek-v4-pro | DeepSeek | ¥0.009 / ¥0.027 | ¥9 / ¥27 | ¥900 / ¥2,700 |
| GLM-5.3glm-5.3 | Zhipu (GLM) | ¥0.008 / ¥0.028 | ¥8 / ¥28 | ¥800 / ¥2,800 |
| GLM-5.3-Flashglm-5.3-flash | Zhipu (GLM) | ¥0.0008 / ¥0.0028 | ¥0.8 / ¥2.8 | ¥80 / ¥280 |
| GLM-5.3-FlashXglm-5.3-flashx | Zhipu (GLM) | ¥0.002 / ¥0.007 | ¥2 / ¥7 | ¥200 / ¥700 |
| GLM-5.1glm-5.1 | Zhipu (GLM) | ¥0.006 / ¥0.024 | ¥6 / ¥24 | ¥600 / ¥2,400 |
| GLM-5glm-5 | Zhipu (GLM) | ¥0.004 / ¥0.018 | ¥4 / ¥18 | ¥400 / ¥1,800 |
| GLM-4.7-Flashglm-4.7-flash | Zhipu (GLM) | ¥0 / ¥0 | ¥0 / ¥0 | ¥0 / ¥0 |
| Qwen3.8-Maxqwen3.8-max | Alibaba Cloud Model Studio (Qwen) | ¥0.012 / ¥0.036 | ¥12 / ¥36 | ¥1,200 / ¥3,600 |
| Qwen3.7-Plusqwen3.7-plus | Alibaba Cloud Model Studio (Qwen) | ¥0.002 / ¥0.008 | ¥2 / ¥8 | ¥200 / ¥800 |
| Qwen-Plusqwen-plus | Alibaba Cloud Model Studio (Qwen) | ¥0.0008 / ¥0.002 | ¥0.8 / ¥2 | ¥80 / ¥200 |
| Qwen3.8-Flashqwen3.8-flash | Alibaba Cloud Model Studio (Qwen) | ¥0.0008 / ¥0.0027 | ¥0.8 / ¥2.7 | ¥80 / ¥270 |
| Qwen-Flashqwen-flash | Alibaba Cloud Model Studio (Qwen) | ¥0.00015 / ¥0.0015 | ¥0.15 / ¥1.5 | ¥15 / ¥150 |
| Qwen3-Coder-Flashqwen3-coder-flash | Alibaba Cloud Model Studio (Qwen) | ¥0.001 / ¥0.004 | ¥1 / ¥4 | ¥100 / ¥400 |
| Kimi K3.1 Prokimi-k3.1-pro | Moonshot AI (Kimi) | ¥0.013 / ¥0.068 | ¥13 / ¥68 | ¥1,300 / ¥6,800 |
| Kimi K3kimi-k3 | Moonshot AI (Kimi) | ¥0.02 / ¥0.1 | ¥20 / ¥100 | ¥2,000 / ¥10,000 |
| Kimi K2.6kimi-k2.6 | Moonshot AI (Kimi) | ¥0.0065 / ¥0.027 | ¥6.5 / ¥27 | ¥650 / ¥2,700 |
| Doubao Seed 2.1 Prodoubao-seed-2-1-pro-260915 | Volcengine Ark (Doubao) | ¥0.006 / ¥0.03 | ¥6 / ¥30 | ¥600 / ¥3,000 |
| Doubao Seed 2.1 Turbodoubao-seed-2-1-turbo-260628 | Volcengine Ark (Doubao) | ¥0.003 / ¥0.015 | ¥3 / ¥15 | ¥300 / ¥1,500 |
| Doubao Seed 2.1 Litedoubao-seed-2-1-lite-260915 | Volcengine Ark (Doubao) | ¥0.0008 / ¥0.0027 | ¥0.8 / ¥2.7 | ¥80 / ¥270 |
| Doubao Seed 2.0 Minidoubao-seed-2-0-mini-260428 | Volcengine Ark (Doubao) | ¥0.0002 / ¥0.002 | ¥0.2 / ¥2 | ¥20 / ¥200 |
| Doubao Seed 2.0 Codedoubao-seed-2-0-code-preview-260215 | Volcengine Ark (Doubao) | ¥0.0032 / ¥0.016 | ¥3.2 / ¥16 | ¥320 / ¥1,600 |
| MiniMax-M3.1-FlashMiniMax-M3.1-Flash | MiniMax | ¥0.001 / ¥0.004 | ¥1 / ¥4 | ¥100 / ¥400 |
| MiniMax-M3MiniMax-M3 | MiniMax | ¥0.0021 / ¥0.0084 | ¥2.1 / ¥8.4 | ¥210 / ¥840 |
| Claude Fable 5.2claude-fable-5-2 | Anthropic (Claude) | $0.008 / $0.04 | $8 / $40 | $800 / $4,000 |
| Claude Fable 5.1claude-fable-5-1 | Anthropic (Claude) | $0.01 / $0.05 | $10 / $50 | $1,000 / $5,000 |
| Claude Opus 5.5claude-opus-5-5 | Anthropic (Claude) | $0.004 / $0.02 | $4 / $20 | $400 / $2,000 |
| Claude Opus 5claude-opus-5 | Anthropic (Claude) | $0.005 / $0.025 | $5 / $25 | $500 / $2,500 |
| Claude Sonnet 5.5claude-sonnet-5-5 | Anthropic (Claude) | $0.002 / $0.01 | $2 / $10 | $200 / $1,000 |
| Claude Sonnet 5claude-sonnet-5 | Anthropic (Claude) | $0.002 / $0.01 | $2 / $10 | $200 / $1,000 |
| Claude Sonnet 4.6claude-sonnet-4-6 | Anthropic (Claude) | $0.003 / $0.015 | $3 / $15 | $300 / $1,500 |
| Claude Haiku 5.5claude-haiku-5-5 | Anthropic (Claude) | $0.0001 / $0.0005 | $0.1 / $0.5 | $10 / $50 |
| Claude Haiku 4.5claude-haiku-4-5-20251001 | Anthropic (Claude) | $0.001 / $0.005 | $1 / $5 | $100 / $500 |
| GPT-6 Astragpt-6-astra | OpenAI (GPT) | $0.01 / $0.05 | $10 / $50 | $1,000 / $5,000 |
| GPT-6.1 Solgpt-6.1-sol | OpenAI (GPT) | $0.002 / $0.01 | $2 / $10 | $200 / $1,000 |
| GPT-6.1 Terragpt-6.1-terra | OpenAI (GPT) | $0.001 / $0.005 | $1 / $5 | $100 / $500 |
| GPT-6 Lunagpt-6-luna | OpenAI (GPT) | $0.0001 / $0.0005 | $0.1 / $0.5 | $10 / $50 |
| GPT-5.5gpt-5.5 | OpenAI (GPT) | $0.005 / $0.03 | $5 / $30 | $500 / $3,000 |
| GPT-5.4gpt-5.4 | OpenAI (GPT) | $0.0025 / $0.015 | $2.5 / $15 | $250 / $1,500 |
| GPT-5.4 minigpt-5.4-mini | OpenAI (GPT) | $0.00075 / $0.0045 | $0.75 / $4.5 | $75 / $450 |
| GPT-5.4 nanogpt-5.4-nano | OpenAI (GPT) | $0.0002 / $0.00125 | $0.2 / $1.25 | $20 / $125 |
| GPT-5.3 Codexgpt-5.3-codex | OpenAI (GPT) | $0.00175 / $0.014 | $1.75 / $14 | $175 / $1,400 |
| Gemini 4 Carbongemini-4-carbon | Google (Gemini) | $0.002 / $0.01 | $2 / $10 | $200 / $1,000 |
| Gemini 4 Argongemini-4-argon | Google (Gemini) | $0.002 / $0.01 | $2 / $10 | $200 / $1,000 |
| Gemini 3.1 Pro Previewgemini-3.1-pro-preview | Google (Gemini) | $0.002 / $0.012 | $2 / $12 | $200 / $1,200 |
| Gemini 3.8 Flashgemini-3.8-flash | Google (Gemini) | $0.00075 / $0.00375 | $0.75 / $3.75 | $75 / $375 |
| Gemini 3.5 Flash-Litegemini-3.5-flash-lite | Google (Gemini) | $0.0003 / $0.0025 | $0.3 / $2.5 | $30 / $250 |
Official price list by provider
Prices are per million tokens in each provider's list currency: standard online prices, excluding batch and off-peak discounts and free tiers. Last checked: 2026-11-12.
DeepSeek
Source: api-docs.deepseek.com/zh-cn/quick_start/pricingPrices shown are peak-hour prices: weekdays (excluding Chinese public holidays) 09:00–12:00 and 14:00–18:00 Beijing time. All other times, including weekends and holidays, are half price. Caching is automatic with no write fee. The old model names deepseek-chat and deepseek-reasoner stopped working on 2026-07-24.
| Model (model ID) | Input | Cache hit | Cache write | Output | Notes |
|---|---|---|---|---|---|
| DeepSeek V4.1 Flashdeepseek-flash | ¥2 | ¥0.04 | — | ¥8 | Peak price; off-peak is half: input ¥1 / cache hit ¥0.02 / output ¥4 |
| DeepSeek V4 Prodeepseek-v4-pro | ¥9 | ¥0.3 | — | ¥27 | Peak price; off-peak is half: input ¥4.5 / cache hit ¥0.15 / output ¥13.5 |
Zhipu (GLM)
Source: docs.bigmodel.cn/cn/guide/start/pricingCache storage is billed hourly and is free for a limited time; it is not included here. Some models are priced by input length and the whole request moves to a higher tier above 32K.
| Model (model ID) | Input | Cache hit | Cache write | Output | Notes |
|---|---|---|---|---|---|
| GLM-5.3glm-5.3 | ¥8 | ¥2 | — | ¥28 | |
| GLM-5.3-Flashglm-5.3-flash | ¥0.8 | ¥0.23 | — | ¥2.8 | Not free. Zhipu's free models are GLM-4.7-Flash and GLM-4-Flash-250414 |
| GLM-5.3-FlashXglm-5.3-flashx | ¥2 | ¥0.57 | — | ¥7 | |
| GLM-5.1glm-5.1 | ¥6 | ¥1.3 | — | ¥24 | Input of 32K or more: input ¥8 / cache hit ¥2 / output ¥28 |
| GLM-5glm-5 | ¥4 | ¥1 | — | ¥18 | Input of 32K or more: input ¥6 / cache hit ¥1.5 / output ¥22 |
| GLM-4.7-Flashglm-4.7-flash | ¥0 | ¥0 | — | ¥0 | Free |
Alibaba Cloud Model Studio (Qwen)
Source: help.aliyun.com/zh/model-studio/model-pricingList prices for the mainland China (Beijing) deployment; current discounts are shown in the Model Studio console. The price list does not give cache prices per model, so cache hits use the implicit-cache rule in Alibaba's docs (20% of input). Qwen3.8 cache prices only appear in the console, so they are counted at the input price here.
| Model (model ID) | Input | Cache hit | Cache write | Output | Notes |
|---|---|---|---|---|---|
| Qwen3.8-Maxqwen3.8-max | ¥12 | — | — | ¥36 | Cache price only appears in the console; the calculator uses the input price |
| Qwen3.7-Plusqwen3.7-plus | ¥2 | ¥0.4 | — | ¥8 | List price; the page shows a limited-time 20% discount. Over 256K input: input ¥6 / cache hit ¥1.2 / output ¥24 |
| Qwen-Plusqwen-plus | ¥0.8 | ¥0.16 | — | ¥2 | Output is the non-thinking price; thinking mode output is ¥8. Higher tiers apply above 128K input |
| Qwen3.8-Flashqwen3.8-flash | ¥0.8 | — | — | ¥2.7 | Cache price only appears in the console; the calculator uses the input price |
| Qwen-Flashqwen-flash | ¥0.15 | ¥0.03 | — | ¥1.5 | 128K–256K input: ¥0.6 / ¥0.12 / ¥6; 256K–1M: ¥1.2 / ¥0.24 / ¥12 (input / cache hit / output) |
| Qwen3-Coder-Flashqwen3-coder-flash | ¥1 | ¥0.2 | — | ¥4 | Higher tiers above 32K input; 32K–128K: input ¥1.5 / cache hit ¥0.3 / output ¥6 |
Moonshot AI (Kimi)
Source: platform.kimi.com/docs/pricing/chatCaching is automatic. 5-minute cache writes cost the input price; 1-hour cache writes cost 2 × input.
| Model (model ID) | Input | Cache hit | Cache write | Output | Notes |
|---|---|---|---|---|---|
| Kimi K3.1 Prokimi-k3.1-pro | ¥13 | ¥1.3 | — | ¥68 | |
| Kimi K3kimi-k3 | ¥20 | ¥2 | — | ¥100 | |
| Kimi K2.6kimi-k2.6 | ¥6.5 | ¥1.1 | — | ¥27 | The K2 series closed to new users on 2026-10-31 and is scheduled to retire at the end of December 2026 |
Volcengine Ark (Doubao)
Source: www.volcengine.com/docs/82379/1544106Standard online inference prices. Cache storage is billed separately at ¥0.017 per million tokens per hour and is not included here. Some models are priced by input-length tier.
| Model (model ID) | Input | Cache hit | Cache write | Output | Notes |
|---|---|---|---|---|---|
| Doubao Seed 2.1 Prodoubao-seed-2-1-pro-260915 | ¥6 | ¥1.2 | — | ¥30 | |
| Doubao Seed 2.1 Turbodoubao-seed-2-1-turbo-260628 | ¥3 | ¥0.6 | — | ¥15 | |
| Doubao Seed 2.1 Litedoubao-seed-2-1-lite-260915 | ¥0.8 | ¥0.16 | — | ¥2.7 | |
| Doubao Seed 2.0 Minidoubao-seed-2-0-mini-260428 | ¥0.2 | ¥0.04 | — | ¥2 | 32K–128K input: ¥0.4 / ¥0.08 / ¥4; 128K–256K: ¥0.8 / ¥0.16 / ¥8 (input / cache hit / output) |
| Doubao Seed 2.0 Codedoubao-seed-2-0-code-preview-260215 | ¥3.2 | ¥0.64 | — | ¥16 | 32K–128K input: ¥4.8 / ¥0.96 / ¥24; 128K–256K: ¥9.6 / ¥1.92 / ¥48 (input / cache hit / output) |
MiniMax
Source: platform.minimaxi.com/docs/pricing/overviewCache hits are billed at the listed price and new input at the input price, with no cache write fee.
| Model (model ID) | Input | Cache hit | Cache write | Output | Notes |
|---|---|---|---|---|---|
| MiniMax-M3.1-FlashMiniMax-M3.1-Flash | ¥1 | ¥0.1 | — | ¥4 | |
| MiniMax-M3MiniMax-M3 | ¥2.1 | ¥0.42 | — | ¥8.4 | Price after the official permanent 50% discount; over 512K input: input ¥4.2 / cache hit ¥0.84 / output ¥16.8 |
Anthropic (Claude)
Source: platform.claude.com/docs/en/about-claude/pricingThe cache write column is the 5-minute cache write price (1.25 × input); 1-hour cache writes cost 2 × input. Cache-hit discounts vary by model; the 5.5 family pays 5% of input.
| Model (model ID) | Input | Cache hit | Cache write | Output | Notes |
|---|---|---|---|---|---|
| Claude Fable 5.2claude-fable-5-2 | $8 | $0.2 | $10 | $40 | |
| Claude Fable 5.1claude-fable-5-1 | $10 | $0.25 | $12.5 | $50 | |
| Claude Opus 5.5claude-opus-5-5 | $4 | $0.2 | $5 | $20 | |
| Claude Opus 5claude-opus-5 | $5 | $0.5 | $6.25 | $25 | |
| Claude Sonnet 5.5claude-sonnet-5-5 | $2 | $0.1 | $2.5 | $10 | |
| Claude Sonnet 5claude-sonnet-5 | $2 | $0.2 | $2.5 | $10 | |
| Claude Sonnet 4.6claude-sonnet-4-6 | $3 | $0.3 | $3.75 | $15 | |
| Claude Haiku 5.5claude-haiku-5-5 | $0.1 | $0.01 | $0.125 | $0.5 | Prompts over 100,000 tokens (cache reads and writes count toward length): input $0.50 / cache hit $0.05 / cache write $0.625 / output $2.50 |
| Claude Haiku 4.5claude-haiku-4-5-20251001 | $1 | $0.1 | $1.25 | $5 |
OpenAI (GPT)
Source: developers.openai.com/api/docs/pricingFor the GPT-6 family, the cache write column is the 30-minute write price; 12-hour writes cost 2.5 × input. Requests with more than 272K input tokens are billed entirely at the long-context price.
| Model (model ID) | Input | Cache hit | Cache write | Output | Notes |
|---|---|---|---|---|---|
| GPT-6 Astragpt-6-astra | $10 | $1 | $12.5 | $50 | Over 272K input: input $20 / cache hit $2 / cache write $25 / output $75 |
| GPT-6.1 Solgpt-6.1-sol | $2 | $0.1 | $2.5 | $10 | Over 272K input: input $4 / cache hit $0.20 / cache write $5 / output $15 |
| GPT-6.1 Terragpt-6.1-terra | $1 | $0.05 | $1.25 | $5 | Over 272K input: input $2 / cache hit $0.10 / cache write $2.50 / output $7.50 |
| GPT-6 Lunagpt-6-luna | $0.1 | $0.01 | $0.125 | $0.5 | Over 272K input: input $0.20 / cache hit $0.02 / cache write $0.25 / output $0.75 |
| GPT-5.5gpt-5.5 | $5 | $0.5 | — | $30 | Over 272K input: input $10 / cache hit $1 / output $45 |
| GPT-5.4gpt-5.4 | $2.5 | $0.25 | — | $15 | Over 272K input: input $5 / cache hit $0.50 / output $22.50 |
| GPT-5.4 minigpt-5.4-mini | $0.75 | $0.075 | — | $4.5 | |
| GPT-5.4 nanogpt-5.4-nano | $0.2 | $0.02 | — | $1.25 | |
| GPT-5.3 Codexgpt-5.3-codex | $1.75 | $0.175 | — | $14 |
Google (Gemini)
Source: aistudio.google.com/docs/pricingPaid-tier prices. Explicit caching also charges an hourly storage fee, not included here; output prices include thinking tokens.
| Model (model ID) | Input | Cache hit | Cache write | Output | Notes |
|---|---|---|---|---|---|
| Gemini 4 Carbongemini-4-carbon | $2 | $0.1 | — | $10 | Rising on 2026-12-01 to input $4 / cache hit $0.20 / output $20 |
| Gemini 4 Argongemini-4-argon | $2 | $0.1 | — | $10 | Rising on 2026-12-01 to input $4 / cache hit $0.20 / output $20 |
| Gemini 3.1 Pro Previewgemini-3.1-pro-preview | $2 | $0.2 | — | $12 | Prompts over 200K tokens: input $4 / cache hit $0.40 / output $18 |
| Gemini 3.8 Flashgemini-3.8-flash | $0.75 | $0.075 | — | $3.75 | Rising on 2027-01-01 to input $1.50 / cache hit $0.15 / output $7.50 |
| Gemini 3.5 Flash-Litegemini-3.5-flash-lite | $0.3 | $0.03 | — | $2.5 |
Getting a more accurate estimate
Enter input, cache and output separately
Each request has three kinds of tokens: regular input, cached input and output, with very different prices. A cache hit is usually 2% to 30% of the input price, and output is usually several times the input price. The usage field in the API response lists them separately (prompt_cache_hit_tokens / prompt_cache_miss_tokens at DeepSeek, cache_read_input_tokens / cache_creation_input_tokens at Anthropic), so copy those numbers in.
Coding agents are mostly cache
Claude Code, Codex and similar tools resend the earlier conversation and file contents every turn, so cached tokens are often several times the new input. Entering only input and output will underestimate the cost, or overestimate it if you count the cached part as regular input.
Long contexts can move you to another price tier
Some providers price by the input length of each request, and the whole request is billed at a higher rate once the input passes a threshold. The calculator uses the lowest tier; check the notes in the price table for long contexts.
USD and CNY
Chinese providers list prices in CNY and others in USD. The calculator converts at the central parity rate on 2026-11-09 ($1 = ¥6.7188); the rate your card or payment channel uses may differ, and you can change it in the calculator.
FAQ
How many words is 1,000 tokens?
DeepSeek's official docs give rough ratios of about 0.6 tokens per Chinese character and 0.3 tokens per English character, so 1,000 tokens is roughly 3,300 English characters. Every model tokenizes differently, so the same text produces different counts at different providers; the usage field in the API response is the real number.
Why does output cost so much more than input?
Almost every provider charges several times more for output than for input, because output is generated one token at a time and uses more compute. Enter input and output separately — a single blended number can be far off.
What is a cache hit, and why is it cheaper?
Multi-turn chats and coding agents resend earlier context with every request. Providers cache the repeated prefix and bill the cached part at a lower cache-hit price, usually 2% to 30% of the input price. Claude and the GPT-6 family charge a separate cache-write price, higher than input, the first time; providers such as DeepSeek and MiniMax charge nothing for writes.
What do 100 million tokens cost?
Multiply the price per million tokens by 100, separately for input and output. The price table on this page already lists every model per 1,000, per million and per 100 million tokens.
How current are these prices?
Every price on this page was last checked against the provider's official price list on 2026-11-12, and each provider links to its official pricing page. Providers change prices and run limited-time offers, so the official page and your bill are the final word.
Are these CCSub's prices?
No. This page lists each provider's public API prices for estimating costs. CCSub has its own prices: input and output are billed at the rates on the models page, and cache reads and writes are billed at the input price rather than with the providers' cache discounts. See the models and pricing pages.
Related pages
See every request's tokens and cost
CCSub's Usage page lists input, cache and output tokens and the charge for every request; CCSub's own prices are on the models page.
