JSON API

Static JSON files, rebuilt from the catalog whenever its data changes; /api/status names the data snapshot being served. No key, no sign-up, no rate limits beyond fair use. Responses carry Access-Control-Allow-Origin: *, an ETag and Cache-Control: public, max-age=300; send If-None-Match to get a 304 when nothing changed.

Endpoints

Path	Contents
`/api.json`	Everything in the models.dev shape: providers → models (keyed by the provider's model id) → cost per 1M tokens, limits, modalities.
`/api/v1/index.json`	Counts, last update, licence and the endpoint list.
`/api/v1/providers.json`	All providers.
`/api/v1/providers/{provider}.json`	One provider and every model it serves, with current prices.
`/api/v1/models.json`	All models, each with its offerings and current prices (exact decimal strings).
`/api/v1/models/{model}.json`	One model with every provider's full, append-only price history and provenance.
`/api/v1/omg-prices.json`	Open Model Gateway Pricing v3 price lines: integer micro-USD per batch, ready to publish.
`/api/v1/history.json`	Every price entry ever recorded, newest first.

Formats

Prices in /api/v1/ are exact decimal strings in USD per unit ("2.5" per 1M tokens), never floats. A missing meter means unknown, not free; "0" is free.
/api.json follows models.dev, so its costs are JSON numbers. The numbers are written exactly as stored (no float rounding). cache_write is the 5-minute write where a provider prices cache TTLs separately; tiers are cost.tiers[].tier.size, applying above that prompt size. It adds catalog_id and has no npm or AI SDK fields.
Prompt-size tiers apply to the whole request when the prompt (all input tokens, cache reads and writes included) is strictly greater than the threshold.
Batch rates appear only when the provider publishes them; they are never derived from standard rates.
omg-prices.json converts each rate to integer micro-USD per Open Model Gateway batch (1M tokens, 1 image, 1,000 ms of audio, ...). A rate that is not a whole micro-USD is rounded up and flagged rounded_up; batch_complete is false when batch lines do not cover the same meters as standard lines, which the gateway requires.
The data-file schema is at /api/v1/schema.json (JSON Schema for the TOML data files).

Example

Fetch the gateway price lines

curl -s https://models.omg.bitop.dev/api/v1/omg-prices.json | jq '.data[0]'

First entry (lines trimmed to two)

{
  "aliases": [],
  "batch_complete": null,
  "batch_price_lines": null,
  "effective_from": "2026-10-09",
  "label": "Global cross-Region inference",
  "limits": {
    "context": 1000000,
    "input": null,
    "output": 128000
  },
  "model": "claude-haiku-5-5",
  "model_name": "Claude Haiku 5.5",
  "price_lines": [
    {
      "batch": 1000000,
      "meter": "input_tokens",
      "microusd_per_batch": "100000",
      "sku_label": "Input",
      "source_usd_per_unit": "0.1",
      "unit_label": "/M tokens"
    },
    {
      "batch": 1000000,
      "meter": "output_tokens",
      "microusd_per_batch": "500000",
      "sku_label": "Output",
      "source_usd_per_unit": "0.5",
      "unit_label": "/M tokens"
    }
  ],
  "provenance": {
    "fetched_at": "2026-10-09T19:52:00Z",
    "kind": "provider-price-list",
    "url": "https://pricing.us-east-1.amazonaws.com/offers/v1.0/aws/AmazonBedrockFoundationModels/current/us-east-1/index.json"
  },
  "provider": "amazon-bedrock",
  "rounded_up": false,
  "upstream_id": "global.anthropic.claude-haiku-5-5"
}