OpenRouter LLM Model Pricing Scraper

Fetches the full OpenRouter model catalog — pricing, context length, capabilities, and supported parameters — as structured JSON. One API call returns everything.

OpenRouter publishes a public endpoint that lists every model it routes to: pricing in USD per token, context windows, supported parameters like tools and structured_outputs, modalities, tokenizer details, and knowledge cutoffs. This actor pulls that data and formats it into clean records you can actually use for model selection, cost estimation, and compatibility filtering.

What It Returns

One record per model. Fields that matter:

Field	Type	Description
model_id	string	Provider/model slug (e.g. openai/gpt-4o)
canonical_slug	string	Versioned slug with datestamp
name	string	Display name
provider	string	Model provider (openai, anthropic, meta-llama, etc.)
context_length	integer	Maximum context window in tokens
max_completion_tokens	integer	Maximum output tokens
modality	string	Input/output modality (e.g. text+image->text)
input_modalities	string	Pipe-delimited list of input types
output_modalities	string	Pipe-delimited list of output types
prompt_price_usd_per_token	number	Price per input token in USD
completion_price_usd_per_token	number	Price per output token in USD
input_cache_read_price_usd_per_token	number	Cached input price per token
input_cache_write_price_usd_per_token	number	Cache write price per token
image_price_usd_per_image	number	Per-image pricing (where applicable)
supported_parameters	string	Pipe-delimited list (temperature, tools, structured_outputs, reasoning, ...)
tokenizer	string	Tokenizer type
instruct_type	string	Fine-tuning format
knowledge_cutoff	string	Training data cutoff date
expiration_date	string	Deprecation date for legacy models
endpoints_url	string	URL to per-provider routing data
profile_url	string	Model page on openrouter.ai
scraped_at	string	ISO timestamp of the run

Most scrapers skip cache-read pricing. This one does not — and cache pricing is the cost lever most teams are actually watching in 2025-26.

Input

Parameter	Type	Default	Description
maxItems	integer	500	Maximum records to return. Leave blank to get all 350+ models.

No API key required. No proxy needed. The OpenRouter catalog is a public endpoint.

Usage

Point it at the defaults and run. The whole catalog comes back in under 5 seconds. Set maxItems if you only need a sample — useful for testing downstream pipelines before a full pull.

{
  "maxItems": 50
}

When to Use This

• Programmatic model selection by capability, context length, or price
• LLM cost estimation and budgeting across providers
• Tracking model deprecations via expiration_date
• Filtering by supported_parameters to find models that support tools, structured_outputs, or reasoning
• Pairing with competitor pricing data for FinOps dashboards

The catalog updates regularly as OpenRouter adds new models. Run it on a schedule to keep your data current.

Data Notes

• Pricing fields are null when a model is free or pricing is not yet set
• supported_parameters lists vary significantly by model — not all models support tools or structured_outputs
• knowledge_cutoff is often null for third-party models where OpenRouter does not publish the information
• Array fields (modalities, parameters) are returned as pipe-delimited strings for flat schema compatibility

---

Questions or issues? Use the feedback fields in the input form or contact us at actor-support@orbtop.com.