Prices verified June 2026 · 45+ models

Know your AI bill before you ship

The most accurate API cost calculator for builders. Compare GPT-5.5, Claude, Gemini and DeepSeek — with prompt caching, batch discounts, reasoning tokens and agent loops modeled in.

Start calculating — free Get Pro — $39.99 once
45Models
13Providers
$0To use
0Logins

Workload presets

Per request

Volume

Cost levers

Batch API (−50%)
Count reasoning tokens

Models to compare

Estimated cost

Model In $/1M Out $/1M $/request $/month $/year

Estimates only. Token counts vary by tokenizer and content. Reasoning models can emit far more output tokens than visible. Always confirm against your provider's invoice.

Token counter & instant cost

Paste any prompt or document. We estimate tokens and price it across your selected models in real time. Heuristic estimate — close to GPT/Claude tokenizers for English.

ModelCost as inputCost as output1K calls/mo
Model Provider Type Input $/1M Output $/1M Cached In Context

Self-hosting

vs API model

Tip: blended ≈ (input price + output price) ÷ 2, weighted to your traffic.

Breakeven analysis

Self-hosting also carries engineering, ops, reliability and idle-capacity costs not shown here. This compares raw compute vs API token spend only.

Your usage

Compare against API model

Subscription vs pay-as-you-go API

PlanProvider$/monthNotes

Every major provider

Live token pricing from official sources. Tap any provider for their docs.

Built for the costs that actually bite

Most calculators stop at input × output. Real AI bills don't.

Caching modeled in

Set your cache hit rate and watch real savings appear. Cached input is ~90% cheaper — the single biggest lever on your bill.

The agent tax

Agents loop. One "request" can be 8+ model calls. The step multiplier shows what autonomous workflows actually cost.

Reasoning tokens

Reasoning models bill for hidden thinking tokens. We flag them so your estimate isn't off by several times.

Batch discounts

One toggle applies the 50% batch-API discount across every provider that offers it.

Shareable & exportable

Every scenario lives in a URL. Share it with your team or export to CSV in one click.

Verified pricing

Numbers checked against official provider pages, dated, and updated as prices change.

How AI API pricing works (2026)

Large language models are billed per token — roughly ¾ of a word in English. You pay separately for input (the prompt and context you send) and output (what the model generates). Output is almost always 3–8× more expensive than input, which is why verbose responses and chatty agents blow up budgets.

monthly cost = ( input_tokens × input_price + output_tokens × output_price ) ÷ 1,000,000 × requests_per_month

The four levers that actually move your bill

1. Prompt caching

If you send the same system prompt, documents, or context repeatedly, providers let you cache it. Cached input typically costs 75–90% less. For RAG, coding assistants and agents — where context is large and stable — caching is the highest-ROI optimization available. Set your realistic cache hit rate in the calculator above to see it.

2. Model selection

Frontier models (Claude Opus 4.8, GPT-5.5) are 10–50× the price of budget tiers (Gemini Flash-Lite, Nova Micro, DeepSeek, Llama 3.1 8B). Most production traffic doesn't need frontier intelligence. Route simple tasks to cheap models and reserve premium models for hard reasoning.

3. Batch processing

If latency doesn't matter — overnight jobs, bulk classification, embeddings — the Batch API cuts cost in half on OpenAI, Anthropic, Google and Amazon. Toggle it on to compare.

4. Output discipline

Because output dominates cost, cap max_tokens, request concise formats, and avoid asking reasoning models to "think out loud" when you don't need it. Reasoning tokens are billed even though you may never see them.

Frequently miscalculated: the agent loop

A single agent "task" can trigger 5–15 model calls as it plans, calls tools, and revises. If you size your budget on one call per task, you'll be off by an order of magnitude. Use the agent steps slider to model this.

Frequently asked questions

Free forever. Pro when you're ready.

Every calculator above is free, no login. Unlock power features once — keep them for life.

Free

$0
  • All 5 calculators
  • 45+ models, 13 providers
  • Caching, batch & agent modeling
  • Share links & CSV export
  • No login, no limits
Start calculating
Best value

Pro — Lifetime

$39.99 once
  • Everything in Free
  • Saved scenarios — save & reload any workload
  • PDF cost reports for stakeholders
  • Full model database CSV export
  • Works on any device with your key
  • Lifetime updates — pay once
Get Lifetime Pro

Pro — Monthly

$4.99 /mo
  • Everything in Pro
  • Cancel anytime
  • Best if you just need it this quarter
Start monthly

Secure checkout by Stripe. Your license key is delivered to your checkout email — if it hasn't landed within a few hours, write coltsinsider@gmail.com and we'll sort it immediately.
30-day money-back guarantee — email us and we'll refund it, no questions asked.
Already purchased? Enter your license key · Manage Pro

Stop guessing what your AI features will cost

Model your real workload above for free. When you need saved scenarios, stakeholder-ready PDF reports and full-database exports, Pro is a single $39.99 purchase — yours for life.

Get Lifetime Pro — $39.99 Keep using it free