aiapicostcalcUnit economicsPricesBrandsCompareBenchmarksSkills & PluginsFind my modelGuidesChanges

Token cost calculator: 1K, 100K and 1M tokens in USD

Short answer: API prices are quoted per 1,000,000 tokens, so 1,000 input tokens cost between $0.0001 and $0.0200 on the 55 verified models here. Output is billed at its own rate and usually dominates a bill: 1,000 output tokens cost between $0.0001 and $0.1000.

Last verified 2026-11-18 · Sources: src/prices.json and the official provider pricing pages it links · Report an error

1
2
Presets

Preset token counts are starting assumptions; replace them with your logged averages.

Cost per call$0.00075
Per 1,000 calls$0.7500

$0.75 in · $4.5 out per 1M

All models, this workload

ModelPer callPer 1K calls
GPT-6 Luna$0.0001$0.1000
Claude Haiku 5.5$0.0001$0.1000
Gemini 2.5 Flash-Lite$0.0001$0.1000
Ministral 3 3B$0.0001$0.1000
GPT-4o mini (deprecated)$0.00015$0.1500
Mistral Small 4$0.00015$0.1500
Ministral 3 8B$0.00015$0.1500
GPT-5.6 Luna$0.0002$0.2000
GPT-5.4 nano (deprecated, retires 2027-04-01)$0.0002$0.2000
Ministral 3 14B$0.0002$0.2000
GPT-5 mini (deprecated, retires 2026-12-11)$0.00025$0.2500
Gemini 3.1 Flash-Lite$0.00025$0.2500
Gemini 3.5 Flash-Lite$0.0003$0.3000
Gemini 2.5 Flash$0.0003$0.3000
DeepSeek Flash (V4.1)$0.0003$0.3000
Codestral 25.08$0.0003$0.3000
Mistral Large 4$0.0007$0.7000
GPT-5.4 mini$0.00075$0.7500
Gemini 3.8 Flash$0.00075$0.7500
Gemini 3.6 Flash$0.00075$0.7500
GPT-6.1 Terra$0.0010$1.00
Claude Haiku 4.5$0.0010$1.00
Grok Build 0.1$0.0010$1.00
o4-mini (shut down 2026-10-23)$0.0011$1.10
Gemini 2.5 Pro$0.00125$1.25
Grok 4.3$0.00125$1.25
DeepSeek V4 Pro$0.00132$1.32
GPT-6.1 Sol$0.0020$2.00
GPT-6 Sol$0.0020$2.00
GPT-5.6 Terra$0.0020$2.00
GPT-4.1 (deprecated)$0.0020$2.00
Claude Sonnet 5.5$0.0020$2.00
Claude Sonnet 5$0.0020$2.00
Gemini 4 Carbon$0.0020$2.00
Gemini 4 Argon$0.0020$2.00
Gemini 3.1 Pro Preview$0.0020$2.00
Grok 4.7$0.0020$2.00
GPT-5.4$0.0025$2.50
Claude Sonnet 4.6$0.0030$3.00
GPT-5.6 Sol$0.0040$4.00
Claude Opus 5.5$0.0040$4.00
GPT-5.5 Rosalind Research$0.0050$5.00
GPT-5.5 Rosalind Discovery$0.0050$5.00
GPT-5.5$0.0050$5.00
Claude Opus 5$0.0050$5.00
Claude Fable 5.2$0.0080$8.00
Claude Mythos 5.2$0.0080$8.00
GPT-6 Astra$0.0100$10.00
GPT-6 Rosalind$0.0100$10.00
Claude Fable 5.1$0.0100$10.00
Claude Mythos 5.1$0.0100$10.00
Claude Fable 5$0.0100$10.00
Claude Mythos 5$0.0100$10.00
GPT-5.6 Cyber$0.0125$12.50
GPT-6 Cyber$0.0200$20.00

o4-mini is shut down; its last published list rate is kept for comparing old invoices, not as a model you can call today — the model page names its replacement.

What a fixed number of tokens costs

Input and output are priced separately. This table applies the standard rate to a fixed token count, so it answers "what does 100,000 tokens cost?" without assuming a workload.

Model1K input100K input1M input1K output1M output
GPT-6 Luna$0.0001$0.0100$0.1000$0.0005$0.5000
Claude Haiku 5.5$0.0001$0.0100$0.1000$0.0005$0.5000
Gemini 2.5 Flash-Lite$0.0001$0.0100$0.1000$0.0004$0.4000
Ministral 3 3B$0.0001$0.0100$0.1000$0.0001$0.1000
GPT-4o mini$0.00015$0.0150$0.1500$0.0006$0.6000
Mistral Small 4$0.00015$0.0150$0.1500$0.0006$0.6000
Ministral 3 8B$0.00015$0.0150$0.1500$0.00015$0.1500
GPT-5.6 Luna$0.0002$0.0200$0.2000$0.0012$1.20
GPT-5.4 nano$0.0002$0.0200$0.2000$0.00125$1.25
Ministral 3 14B$0.0002$0.0200$0.2000$0.0002$0.2000
GPT-5 mini$0.00025$0.0250$0.2500$0.0020$2.00
Gemini 3.1 Flash-Lite$0.00025$0.0250$0.2500$0.0015$1.50
Gemini 3.5 Flash-Lite$0.0003$0.0300$0.3000$0.0025$2.50
Gemini 2.5 Flash$0.0003$0.0300$0.3000$0.0025$2.50
DeepSeek Flash (V4.1)$0.0003$0.0300$0.3000$0.0012$1.20
Codestral 25.08$0.0003$0.0300$0.3000$0.0009$0.9000
Mistral Large 4$0.0007$0.0700$0.7000$0.0021$2.10
GPT-5.4 mini$0.00075$0.0750$0.7500$0.0045$4.50
Gemini 3.8 Flash$0.00075$0.0750$0.7500$0.00375$3.75
Gemini 3.6 Flash$0.00075$0.0750$0.7500$0.00375$3.75
GPT-6.1 Terra$0.0010$0.1000$1.00$0.0050$5.00
Claude Haiku 4.5$0.0010$0.1000$1.00$0.0050$5.00
Grok Build 0.1$0.0010$0.1000$1.00$0.0020$2.00
o4-mini$0.0011$0.1100$1.10$0.0044$4.40
Gemini 2.5 Pro$0.00125$0.1250$1.25$0.0100$10.00
Grok 4.3$0.00125$0.1250$1.25$0.0025$2.50
DeepSeek V4 Pro$0.00132$0.1320$1.32$0.00396$3.96
GPT-6.1 Sol$0.0020$0.2000$2.00$0.0100$10.00
GPT-6 Sol$0.0020$0.2000$2.00$0.0100$10.00
GPT-5.6 Terra$0.0020$0.2000$2.00$0.0120$12.00
GPT-4.1$0.0020$0.2000$2.00$0.0080$8.00
Claude Sonnet 5.5$0.0020$0.2000$2.00$0.0100$10.00
Claude Sonnet 5$0.0020$0.2000$2.00$0.0100$10.00
Gemini 4 Carbon$0.0020$0.2000$2.00$0.0100$10.00
Gemini 4 Argon$0.0020$0.2000$2.00$0.0100$10.00
Gemini 3.1 Pro Preview$0.0020$0.2000$2.00$0.0120$12.00
Grok 4.7$0.0020$0.2000$2.00$0.0060$6.00
GPT-5.4$0.0025$0.2500$2.50$0.0150$15.00
Claude Sonnet 4.6$0.0030$0.3000$3.00$0.0150$15.00
GPT-5.6 Sol$0.0040$0.4000$4.00$0.0200$20.00
Claude Opus 5.5$0.0040$0.4000$4.00$0.0200$20.00
GPT-5.5 Rosalind Research$0.0050$0.5000$5.00$0.0250$25.00
GPT-5.5 Rosalind Discovery$0.0050$0.5000$5.00$0.0250$25.00
GPT-5.5$0.0050$0.5000$5.00$0.0300$30.00
Claude Opus 5$0.0050$0.5000$5.00$0.0250$25.00
Claude Fable 5.2$0.0080$0.8000$8.00$0.0400$40.00
Claude Mythos 5.2$0.0080$0.8000$8.00$0.0400$40.00
GPT-6 Astra$0.0100$1.00$10.00$0.0500$50.00
GPT-6 Rosalind$0.0100$1.00$10.00$0.0500$50.00
Claude Fable 5.1$0.0100$1.00$10.00$0.0500$50.00
Claude Mythos 5.1$0.0100$1.00$10.00$0.0500$50.00
Claude Fable 5$0.0100$1.00$10.00$0.0500$50.00
Claude Mythos 5$0.0100$1.00$10.00$0.0500$50.00
GPT-5.6 Cyber$0.0125$1.25$12.50$0.0750$75.00
GPT-6 Cyber$0.0200$2.00$20.00$0.1000$100

Rows are ordered by the cost of 1M input tokens. Cached-input and batch rates are lower where the provider publishes them; the check date for each row is on its model page.

How a token count becomes a price

Total cost = (input tokens x input rate + output tokens x output rate) / 1,000,000, using the rates on the provider’s own pricing page. Because every rate is per million tokens, divide by 1,000 for the cost of 1,000 tokens: a model at $2 per 1M input tokens charges $0.002 for 1,000 input tokens.

Reading rates without a calculator

At $2 per 1M tokens a single token is $0.000002, which is why small volumes look free and monthly volume decides the bill. 10,000 calls with 1,000 input tokens each is 10,000,000 input tokens, or $20 at that rate before output is counted.

Turning text into a token count

Tokenizers differ by model, but for planning: plain English is roughly four characters per token (about 750 words per 1,000 tokens); code, JSON and URLs take more tokens per word because punctuation and camelCase split into chunks; Chinese, Japanese and Korean text takes fewer characters per token. The token estimation guide shows how to get the exact count from the provider tokenizer or from your own request logs.

Worked examples with the current rates

1,000-token prompt and 500-token answer on GPT-5.4 mini

1,000 input tokens cost $0.00075, 500 output tokens cost $0.00225, and the whole call costs $0.0030. At 10,000 calls a month the token bill is $30.00; at 100,000 calls it is $300.

1,000-token prompt and 500-token answer on Claude Sonnet 5.5

1,000 input tokens cost $0.0020, 500 output tokens cost $0.0050, and the whole call costs $0.0070. At 10,000 calls a month the token bill is $70.00; at 100,000 calls it is $700.

1,000-token prompt and 500-token answer on DeepSeek Flash (V4.1)

1,000 input tokens cost $0.0003, 500 output tokens cost $0.0006, and the whole call costs $0.0009. At 10,000 calls a month the token bill is $9.00; at 100,000 calls it is $90.00.

The calculator at the top of this page runs the same arithmetic; change the model or either token count and every figure follows. To turn a monthly bill into a per-user price, use the unit-economics calculator. If those tokens come from an AI coding agent, a workflow tool or an MCP server rather than a plain prompt, the AI tool token cost tables show what each one burns per task.

Video generation is billed in its own units: the Google Flow credit cost calculator converts Flow credits per second and per clip into dollars for each Google AI plan. Images are billed per image instead of per token: the Nano Banana Pro API cost calculator turns 1,120-token 1K/2K images and 2,000-token 4K images into dollars. Google’s open Gemma models are priced on their own two routes: Gemma API pricing covers the free Gemini API tier and the per-token Vertex AI rate.

What AI agents, skills and tools cost in tokens

Each tool page prices three real token scenarios against the same official model rates as the rest of this site, and says whether the token figure is documented, measured or estimated.

Browse models by brand: models by brand · price tiers

Questions

How many tokens is 1,000 words?

Plain English runs at roughly four characters per token, so 1,000 tokens is about 750 words and 1,000 words is roughly 1,300-1,400 tokens. Code, JSON and URLs take more tokens per word because punctuation and symbol names split into chunks; Chinese, Japanese and Korean text takes fewer characters per token. Every tokenizer splits text differently, so use these as planning figures and take the exact count from the model's own tokenizer or from your logs.

Why do output tokens cost more than input tokens?

Generating text one token at a time is more compute-intensive than reading a prompt, so providers price output higher. Across the 55 verified models here the output-to-input ratio runs from 1.0x to 8x, which is why a long answer can cost more than a long prompt on the same model.

Does the calculator include cached input, batch or long-context prices?

It applies the standard list rate, plus the cached-input rate when you enter a cached % and the provider publishes one. Batch discounts, long-context tiers and committed-use rates are not switched on automatically; each model page lists the tiers, and the caching and batch guides explain them.

Which model is cheapest per token?

On 1,000 input tokens the cheapest verified model is GPT-6 Luna at $0.0001. Input and output are priced separately, so the cheapest model for a prompt-heavy job is not always the cheapest for a long answer; the table above is ordered by the cost of 1M input tokens.

Where do these prices come from?

Each rate is copied by hand from the provider's own pricing page and carries a check date; the source URL and date are on every model page. The rules, including how unverified rows are excluded, are on the methodology page.

This page starts from a token count. If you have the text and need the count first, the token calculator counts pasted text for five model families and prices the same text per send and per month.

Next step → Price a whole workload per user

Price data last checked 2026-11-18 against official sources · how we verify

Price-change alerts

One short email when a provider changes a price or a promo ends. No spam, unsubscribe any time.