Read-only model data, no API key.

GET /api/v1/models?q=claude&provider=Anthropic&sort=release&page=1&limit=50

Filters: q, provider, modality, context, and max_price.

Sort values: release, benchmark-coverage, updated, name, context, input-price, and output-price.

Use benchmark=Benchmark Name|Version to rank compatible results, or fields=slug,name,speed_tps to return only selected fields. Pagination is controlled with page and limit; the maximum limit is 200.

GET /api/v1/models?sort=benchmark-coverage&fields=slug,name,intelligence_score,coding_score,agentic_score&limit=10

The response contains a data array and meta with page, limit, and total.

GET /api/v1/models/{provider}/{model}

Returns the model specification, prices, context and output limits, modalities, benchmark data, hosting offers, source links, and verification metadata. Missing slugs return a JSON 404 response.

Available model fields

Use these keys with ?fields= on the list endpoint:

slug, name, developer, description, context_length, max_output_tokens, modalities, release_date, open_weights, input_price_per_million, output_price_per_million, cached_input_price_per_million, intelligence_score, coding_score, agentic_score, speed_tps, benchmark_data, metadata, offerings, source, source_url, verification_status, confidence, and last_verified.

Missing values: undisclosed prices are returned as null. A numeric 0 is reserved for models explicitly listed as free; ingestion sentinels are never exposed by the public API or CSV.

Usage

Responses are publicly cached and do not require authentication. Apply reasonable request rates, preserve ModelBench attribution, and keep original source URLs when reusing the data.