LLM pricing

Free LLM pricing API

In short: GET https://fiscallm.com/api/v1/prices returns the current list price of every model in our price table as JSON. No key. Prices were last checked on 2026-10-18.

List prices

curl "https://fiscallm.com/api/v1/prices?provider=anthropic&q=haiku"

Both filters are optional: provider is openai, anthropic or google, and q matches part of a model ID. The answer has count and models, plus the currency, unit and the date of the last check.

One model

curl https://fiscallm.com/api/v1/prices/claude-sonnet-4-5

{
  "currency": "USD",
  "unit": "per 1,000,000 tokens (unitPrices are per image, second, minute or character)",
  "tier": "list price, standard tier",
  "checked": "2026-10-18",
  "source": "https://fiscallm.com/llm-pricing",
  "audit": "https://github.com/magna56/fiscallm/blob/main/ACCURACY.md",
  "license": "MIT",
  "model": {
    "id": "claude-sonnet-4-5",
    "provider": "anthropic",
    "inputPerMtok": 3,
    "outputPerMtok": 15,
    "cachedInputPerMtok": 0.3,
    "cacheWritePerMtok": 3.75,
    "url": "https://fiscallm.com/llm-pricing/claude-sonnet-4-5"
  }
}

Dated IDs and provider prefixes resolve to the same model, so openai/gpt-4o-mini works. Where they apply, a model also has longContext (the rate above a prompt size), unitPrices (images, audio, video) and next (an announced price change and its date). An unknown model returns 404.

Estimate a call

curl "https://fiscallm.com/api/v1/estimate?model=gpt-4o-mini&input=1500&output=400&cached=1000&calls=300"

{
  "provider": "openai",
  "model": "gpt-4o-mini",
  "costUsd": 0.00039,
  "calls": 300,
  "totalUsd": 0.117,
  "inputTokens": 1500,
  "outputTokens": 400,
  "cachedTokens": 1000,
  "cacheWriteTokens": 0,
  "currency": "USD",
  "checked": "2026-10-18"
}

For AI agents

The same data is available to assistants as tools through the FiscaLLM MCP server at https://fiscallm.com/api/mcp, which adds model comparison and subscription break-even. The API is described in openapi.json.

From list price to your real cost

A price table tells you what a call costs. It does not tell you which of your users cost more than they pay. FiscaLLM adds up real usage per user and sets it next to revenue. See the demo or read how it works.

Questions