Models OpenAI/Anthropic cached-prompt savings (typical hit rate)
Models 0 selected · max 12Direct catalog + 370+ via OpenRouter
Estimated monthly cost
Pick at least one model to see costs.
Model
Provider
Input $/M
Output $/M
Per request
Monthly
Yearly
Flat plan break-even
Flat plans are rate-limited chat products, not raw API access. Use this as a sanity check, not an apples-to-apples comparison.
Pick 1–12 models above to see estimated monthly costs.
Use the workload sliders to size your traffic. Defaults model a small product with 10k API requests per month.
🔒 Workload templates + monthly forecasting
Save named workloads, track real spend over time vs forecast, alert when projection overshoots budget. For ops/finance teams managing multi-team LLM budgets.
AI Cost Calculator — Estimate Monthly LLM API Spend Across GPT, Claude, Gemini and 370+ Models
freeaicostcalculator.app is a free, browser-based calculator that estimates your monthly LLM API spend before you commit. You enter three numbers — requests per month, average input tokens per request, average output tokens per request — pick which models to compare from a curated catalog of 24 models across 8 providers (plus the entire OpenRouter live catalog of 370+ more), and immediately see per-model monthly cost, a horizontal bar-chart comparison, and a flat-plan break-even check against ChatGPT Plus, Claude Pro, Cursor and GitHub Copilot.
This is a planning and budgeting tool. Different from a token counter, which answers "how much does this specific text cost?" — freeaicostcalculator.app answers "if I run this workload at scale, what is my monthly bill across these models?" The calculation is pure arithmetic on each provider's published per-million pricing: per-request cost equals (input tokens × input rate + output tokens × output rate) divided by one million; monthly cost equals per-request cost multiplied by your requests-per-month input.
How does freeaicostcalculator.app work?
Drag the three sliders or type values into the workload panel. Pick 1 to 12 models from the picker — the OpenRouter tab loads the full live catalog (371+ models) on page load with pricing pulled inline, plus a provider sub-filter dropdown (OpenAI, Anthropic, Google, Meta/Llama, Mistral, DeepSeek, xAI, Cohere, Qwen, Perplexity, NVIDIA, MiniMax and 50 more) and a search box. The results panel updates instantly with a sorted bar chart, a detailed table, and a flat-plan break-even card. Toggle the "prompt cache" checkbox to model a 50% effective discount on input cost, which approximates what most production deployments actually pay.
How is this different from a token counter?
freetokencounter.app answers a per-prompt question: paste text, see tokens, see cost for one call. freeaicostcalculator.app answers a workload question: at N requests per month with M tokens average, what is the monthly bill across each model? Different intent — token counting is what you do when iterating on a prompt; cost calculation is what you do when sizing a budget or picking a model for a product. The two tools cross-link naturally: count tokens on a representative prompt with the token counter, then plug those numbers into the cost calculator to project monthly spend.
Should I use a flat plan instead of paying API rates?
It depends on usage. ChatGPT Plus is $20 per month, ChatGPT Pro is $200 per month, Claude Pro is $20 per month, Claude Max is $100 or $200 per month, Cursor Pro is $20 per month, GitHub Copilot is $10 to $19 per month. For light usage, a flat plan is almost always cheaper. For programmatic access at scale, the API is the only option because flat plans are rate-limited chat products without programmatic guarantees. freeaicostcalculator.app shows your projected API spend alongside flat-plan prices so you can spot the break-even point — but treat the comparison as informational, not direct.
AI Model Pricing Comparison Matrix
Model
Provider
Context
Input $/M
Output $/M
Claude Opus 4.7
Anthropic
1M
$15
$75
Claude Sonnet 4.6
Anthropic
1M
$3
$15
Claude Haiku 4.5
Anthropic
200K
$1
$5
GPT-5
OpenAI (via OpenRouter)
400K
$1.25
$10
GPT-5 mini
OpenAI (via OpenRouter)
400K
$0.25
$2
GPT-4.1
OpenAI (via OpenRouter)
1M
$3
$12
Gemini 2.5 Pro
Google
1M
$1.25
$10
Gemini 2.5 Flash
Google
1M
$0.30
$2.50
Grok 4
xAI
256K
$5
$15
Grok 4 Fast
xAI
2M
$0.20
$0.50
DeepSeek V3.1
DeepSeek
128K
$0.27
$1.10
Llama 3.3 70B (Groq)
Groq
128K
$0.59
$0.79
Llama 3.1 8B (Groq)
Groq
128K
$0.05
$0.08
Mistral Large 2
Mistral
128K
$2
$6
Command A
Cohere
256K
$2.50
$10
Prices as of 2026-05-04. Always verify against each provider's pricing page for production budgeting.
Why use freeaicostcalculator.app?
Workload-driven
Three numbers: requests, input tokens, output tokens. No fluff.
Flat-plan break-even
Compare against ChatGPT Plus, Claude Pro, Cursor and GitHub Copilot.
370+ models live
OpenRouter's full catalog with provider sub-filter and search.
No keys, no signup
Pure arithmetic. The only network call is the public OpenRouter catalog.
Frequently Asked Questions
What is freeaicostcalculator.app?
freeaicostcalculator.app is a free, browser-based calculator that estimates your monthly LLM API spend across multiple AI models. You enter three numbers — requests per month, average input tokens per request, average output tokens per request — and pick which models to compare. The calculator shows per-model monthly cost, a horizontal bar chart, and a flat-plan break-even check against ChatGPT Plus, Claude Pro, Cursor and GitHub Copilot.
How does freeaicostcalculator.app calculate cost?
Per-request cost equals (input tokens × input rate + output tokens × output rate) divided by one million, using each model's published per-million pricing. Monthly cost equals per-request cost multiplied by your requests-per-month input. The calculator does not call any AI provider — it is pure arithmetic on baked-in pricing data plus the live OpenRouter catalog fetched at page load.
How is this different from freetokencounter.app?
freetokencounter.app answers the question "how much does this specific text cost when sent to model X?" — you paste a prompt and see its tokens and the cost of that single call. freeaicostcalculator.app answers a different question: "if I run a workload of N requests per month at average M tokens per request, what is my total monthly bill across these models?" One is for a specific prompt in your hand; the other is for planning and budgeting before you have a prompt.
Which models does freeaicostcalculator.app support?
The curated catalog covers Anthropic (Claude Opus 4.7, Sonnet 4.6, Haiku 4.5), Google (Gemini 2.5 Pro, 2.5 Flash, 2.0 Flash), xAI (Grok 4, Grok 4 Fast, Grok 3), Mistral (Large 2, Medium 3, Small 3.1), DeepSeek (V3.1, R1), Groq-hosted Llama (Llama 4 Scout 17B, 3.3 70B, 3.1 8B), Cohere (Command A, R+, R) and OpenAI via OpenRouter (GPT-5, GPT-5 mini, GPT-4.1, o4-mini). On top of that, the OpenRouter tab loads the entire live catalog — 370+ models with pricing pulled directly from OpenRouter at page load.
Should I use a flat plan like ChatGPT Plus instead of paying API rates?
It depends on usage. ChatGPT Plus is $20 per month, ChatGPT Pro is $200 per month, Claude Pro is $20 per month, Claude Max is $100 or $200 per month, Cursor Pro is $20 per month, GitHub Copilot is $10 to $19 per month. freeaicostcalculator.app shows your projected API spend alongside these flat plans so you can spot break-even points. The honest caveat: flat plans are rate-limited chat products, not raw API access — they are not directly comparable when you need full programmatic access for a product or pipeline.
Are the prices accurate?
freeaicostcalculator.app uses each provider's published per-million input and output rates baked into a static pricing object, plus the live OpenRouter catalog fetched on page load. Prices change occasionally; for production budgeting always verify against the provider's current pricing page. The pricing data is dated in the status bar so you can see how fresh it is.
Does freeaicostcalculator.app need an API key?
No. freeaicostcalculator.app does not call any AI provider — it is a pure calculator. The only network request is a public, unauthenticated GET to OpenRouter's models endpoint (no key required) to populate the live catalog. No personal data is sent anywhere.
Can I model prompt caching?
Yes. freeaicostcalculator.app has an optional toggle that applies a 50% effective discount on the input portion of cost, modeling a typical cached-prompt workload. It is a simplification — actual cache savings depend on hit rate and provider (OpenAI typically 50%, Anthropic prompt cache up to 90% on cached portion). For precise modeling, run the calculation twice with different multipliers and split your traffic accordingly.