79 models · 11 providers · checked 2026-07-02

Know exactly what AI costs — before the invoice tells you.

Meterlark tracks published API pricing for every major model, continuously, so "what will this cost" has a real answer instead of a guess. The directory and calculators are free forever. When you want to see what you're actually spending, the workspace is $19/month.

Meter panel

USD / 1M tokens
  • Amazon Nova 2.0 ProAmazon (Nova)$1.38 in · $11.00 out
  • Claude Fable 5Anthropic$10.00 in · $50.00 out
  • DeepSeek V4 FlashDeepSeek$0.14 in · $0.28 out
  • Gemini 2.5 Flash-LiteGoogle$0.10 in · $0.40 out
  • Llama 4 MaverickMeta$0.15 in · $0.60 out
  • Mistral Large 3Mistral AI$0.50 in · $1.50 out

One workload, three price tags

2K in · 500 out per request · 10K requests/day

Claude Fable 5

$13,698

per month at this volume

GPT-5.4 mini

$1,142

per month at this volume

Llama 4 Scout

$106.54

per month at this volume

Example workload, not a recommendation — the priciest model here costs 129x the cheapest for the same volume. Run your own numbers on the calculator.

The problem

Your AI spend lives in six different consoles, and none of them talk to each other.

OpenAI's usage page shows OpenAI. Anthropic's shows Anthropic. Whatever you're paying Google, Mistral or whoever else lives wherever they decided to put it. Nobody hands you the one number that actually matters: what AI cost this month, total, across every model your team touches.

Prices also move without much warning — a provider cuts input pricing, adds a cache tier, or ships a new flagship that quietly makes last quarter's cost model wrong. Catching that requires someone to keep checking every provider's pricing page by hand, which nobody actually has time to do.

The directory

79 models, one page, prices that are actually current.

Every model gets checked against its provider's own pricing page, not scraped once and left to rot. Anything we couldn't independently verify is labelled as such instead of quietly presented as fact.

Browse the full directory

Example, not your data

Model

Amazon Nova 2.0 Pro

Tokens

50K in / 12K out

Estimated cost$0.2008

Calculators

"What will this cost" answered in one page load.

Put in a token volume and request rate, get an exact cost from live pricing — no spreadsheet, no digging through a provider's tier structure to find the number you actually need.

Open the calculator

The workspace

The directory tells you what AI costs. The workspace tells you what it's costing you.

Create a project per product, team or client. Log usage by hand, or drop in a CSV export from a provider's dashboard — either way, cost is computed from the same verified pricing you see in the directory, not whatever the provider chose to invoice you.

See Pro pricing

Spend this month

Example workspace · priced on Amazon Nova 2.0 Pro

$2,235

Support bot$602.71
Content pipeline$502.26
Internal search$1,130

Budgets & alerts

Example workspace · Pro feature

  • Support bot$602.71 / $450.00

    Overalert sent at 134% of budget

  • Content pipeline$502.26 / $520.00

    Nearingalert sent at 97% of budget

  • Internal search$1,130 / $700.00

    Overalert sent at 161% of budget

What if we switched?

Example workspace · same month-to-date usage, every model

  • Claude Haiku 4.5Cheapest$2.00
  • Amazon Nova 2.0 Pro$3.66
  • Claude Sonnet 5Current$4.00
  • Claude Opus 4.8$10.00
  • Claude Fable 5$20.00

Built for agents too

A pricing directory an agent can actually read.

Every model in the directory — all 79 of them — is available as plain JSON through the public API, plus an llms.txt summary for agents that just need the shape of the dataset before they call it. No auth for read access, no scraping.

Read the API docs
GET /api/v1/models/claude-fable-5

{
  "slug": "claude-fable-5",
  "provider": "anthropic",
  "inputPerMTok": 10,
  "outputPerMTok": 50,
  "verifiedAt": "2026-07-02"
}

Prices checked as of 2026-07-02. Check yours next.

The directory and calculators stay free, no account needed. Bring in your real usage whenever you're ready to see the total.