AI API pricing calculator
Pick any tracked model, enter how many tokens a typical request sends and receives, and set your monthly request volume. The estimate uses the same list prices shown in the pricing database, which refresh every six hours.
232 models tracked · Prices updated Oct 27, 2026
Estimated monthly cost
$61.75
- Cost per request
- $0.0012
- Input tokens
- $18.00
- Output tokens
- $43.75
- Annualised
- $741.00
Estimates use public list pricing per 1M tokens. Batch, enterprise, and regional rates may differ.
How to estimate API cost
- 01
Measure one real request
Take a representative prompt, including the system prompt, retrieved context and conversation history, and note its input tokens. Do the same for a typical response. If you only have text, the Token Calculator converts words to an approximate token count.
- 02
Keep input and output separate
Almost every provider charges more for output than for input, often four to eight times more. A single blended rate hides that, so long answers make a cheap-looking model expensive.
- 03
Multiply by realistic volume
Requests per month is usually active users times requests per user. Count retries, tool calls and background jobs too, because each one is billed as a separate request.
- 04
Apply caching only where it happens
Cached input is billed at a lower rate on providers that support it, but only the repeated part of a prompt qualifies, such as a fixed system prompt. Use the share you see in your usage logs rather than a hopeful guess.
Worked example
A support assistant at 50,000 requests a month
Each request sends 1,200 input tokens and returns 350 output tokens, which is 60 million input and 17.5 million output tokens a month. At an illustrative rate of $1.00 per million input and $4.00 per million output, that comes to $60 for input plus $70 for output, or $130 a month. If half the input is a cached system prompt billed at $0.10 per million, input falls to $33 and the total to $103. These rates are examples, not a quote for a named model, so check the live rate for the model you plan to use.
Pick the right calculator
The calculator above covers a single model and workload. For other questions, these tools use the same pricing data.
- Which model is cheapest for my workload?Cheapest LLM API Finder: Rank the 10 lowest estimated token costs for your workload.
- I have text, not token countsToken Calculator: Estimate token counts from raw text and see what it costs on any model.
- Will my prompt and documents fit?Context Window Calculator: See how much conversation, document, or code fits in a model's context window.
- I only use OpenAI modelsGPT API Cost Calculator: Estimate OpenAI GPT API costs by token volume, request count, and model.
- I only use Claude modelsClaude Cost Calculator: Model Anthropic Claude API spend across Opus, Sonnet, and Haiku.
- I only use Gemini modelsGemini Cost Calculator: Estimate Google Gemini API costs including 1M-token context workloads.
- I route calls through OpenRouterOpenRouter Cost Calculator: Compare routed model costs across every provider in one estimate.
- We pay per seat for AI toolsAI Subscription Cost Calculator: Total your team's AI seat spend across every tool and plan.
- Is the spend worth the time saved?AI ROI Calculator: Weigh hours saved and labour cost against monthly AI spend.
What the estimate leaves out
- Batch, enterprise, committed-use and regional rates are not included. They are often lower than list price.
- Image, audio, search and tool-use fees are billed separately on many providers and are not part of a token estimate.
- Tokenizers differ between model families, so the same text can produce a different token count on each.
- Nothing you enter is stored or sent to a provider. The calculation runs in your browser.
Frequently asked questions
- How do I calculate AI API cost?
- Multiply input tokens per request by the input price per million tokens, do the same for output tokens and the output price, add the two, then multiply by requests per month. The calculator on this page does that for any tracked model.
- Where do the prices come from?
- From providers' published list prices, collected through a pricing feed that syncs every six hours. Each model in the pricing database shows when its record was last updated.
- Why is my invoice different from the estimate?
- Invoices reflect actual token counts, retries, cached input, batch discounts and any non-token fees. Compare the estimate with a week of real usage logs and adjust the inputs to match.
- Is the cheapest model the right choice?
- Not always. Price is one factor alongside accuracy, speed, rate limits and data handling terms. Test two or three candidates on real prompts before switching.