OpenRouter AI Model Pricing Scraper

Every model OpenRouter lists — anthropic/claude-*, openai/*, google/*, open-weight models and the rest — with prompt and completion price per million tokens, cache price, context length, max output, input/output modalities and supported parameters.

Try it now — the input is already filled in. Press Start: the 50 newest models on OpenRouter, with price per million tokens and context length, in about 7 seconds, for about $0.11.

name	promptPricePerMTok	completionPricePerMTok	contextLength
Elorian: Glimpse	0.6	3	262144
Meta: Muse Spark Open	1.25	4.25	1048576
Google: Gemini 4 Argon	2	10	1048576

Run it again and each model whose price changed in between says so. Two of our runs 78 minutes apart on 17 Oct 2026 caught DeepSeek Flash Latest going from $0.032 to $0.04 (prompt) and from $0.2 to $0.7 (completion) per million tokens — on that model's row as changes: [{ field, from, to }]. Models that disappear come back as removed rows.

Run it on a schedule and it tells you exactly what changed since the last run: new models, removed models, and every price or context change as { field, from, to }.

Use it for

• LLM cost tracking — alert when a model you use changes price
• Model launch monitoring — see new models the day they appear
• Price comparison tables for newsletters, dashboards and internal tools

Notes

• Prices are converted from OpenRouter's per-token strings to dollars per million tokens (the unit everyone quotes).
• Router models with variable pricing (e.g. openrouter/auto) report -1 upstream; that becomes null, never 0. isFree is true only when both prices are exactly 0.

Example output

One real row from a test run (long text shortened here with …). On later runs changes lists what moved, as { field, from, to }.

{
  "type": "item",
  "source": "openrouter",
  "modelId": "perceptron/perceptron-mk1.5",
  "name": "Perceptron: Perceptron Mk1.5",
  "provider": "perceptron",
  "createdAt": "2026-09-25T16:11:01.000Z",
  "contextLength": 36864,
  "maxCompletionTokens": 8192,
  "inputModalities": [
    "text",
    "image",
    "video",
    "audio"
  ],
  "outputModalities": [
    "text"
  ],
  "promptPricePerMTok": 0.15,
  "completionPricePerMTok": 1.5,
  "cacheReadPricePerMTok": null,
  "webSearchPricePerCall": null,
  "isFree": false,
  "supportedParameters": [
    "frequency_penalty",
    "include_reasoning",
    "max_tokens",
    "presence_penalty",
    "reasoning",
    "reasoning_effort",
    "…"
  ],
  "knowledgeCutoff": null,
  "expirationDate": null,
  "huggingFaceId": null,
  "description": "Perceptron Mk1.5 is Perceptron's embodied reasoning model for physical agents. It accepts text, image, video, and audio input, and answers with text plus option…",
  "url": "https://openrouter.ai/perceptron/perceptron-mk1.5",
  "isNew": null,
  "changes": []
}

Get only what changed, every morning (3 steps)

1. Open this Actor in Apify Console, set What to return to Only what is new or changed since the last run, and click Save as a new task.
2. On the task, add a Schedule — for example daily at 07:00.
3. Open the task's Integrations tab and connect Slack, email, Google Sheets (or Zapier / Make / a webhook). Each morning you get only the new, changed and removed entries.

You pay only for the rows returned plus $0.01 per run, so a quiet day costs about a cent.

Pricing

$2.00 per 1,000 entries, plus $0.01 per run. Removed-entry rows and errors are free. Use Max entries and Only what is new or changed to keep a scheduled run cheap.

**Measured, not estimated:** about **$0.0004 per 1,000 entries** of Apify platform usage at 512 MB (our running cost — included in the price, nothing on top) (458 entries in 30 s, 2026-09-27).

For AI agents

Rows have type: item, removed or error. isNew: null means "no previous run to compare with", not "old". changes is a list of { field, from, to }.

More from northbell

Job boards with exact new/closed tracking: Greenhouse · Lever · Ashby · Workday · AI companies.