$ AICostSaverbeta
Sign in

API reference

Public REST API for LLMs, agents and scripts. Free, read-only, no API key required.

Quickstart

# cheapest offers for a model
curl -s "https://costsaver.coopermed.beget.tech/api/v1/pricing?model=gpt-4o&sort=price_input&limit=5"

# what a specific model costs at every provider
curl -s "https://costsaver.coopermed.beget.tech/api/v1/models/deepseek-v3-2"

The canonical base path is /api/v1; /api is an alias of v1 (same endpoints, e.g. /api/providers). Bare /api returns a JSON index of the API. Interactive playground: /docs, machine schema: /openapi.json.

Conventions

  • All endpoints are GET and return JSON. No authentication; responses are safe to cache.
  • List endpoints return {"items": [...], "meta": {"total", "limit", "offset"}}; detail endpoints return a single object.
  • Money is a string (lossless Decimal): price_input_per_mtok / price_output_per_mtok are USD per 1M tokens, usd_per_hour is USD per hour.
  • checked_at / last_synced_at are ISO 8601 timestamps — use them to judge data freshness.
  • is_available = the provider currently lists this offer.
  • Rate limit: per client IP, answered with 429 and a Retry-After header. Fair use — please cache.

GET /api/v1/providers

All active providers in the catalog.

Param Default Meaning
q — substring match on provider name/slug
limit · offset 100 · 0 pagination (limit ≤ 500); total count is in meta.total

GET /api/v1/providers/{slug}

One provider plus aggregate counts of its catalog entries (models_count, gpu_offers_count, agent_plans_count) and last_synced_at.

GET /api/v1/models

All models with per-model aggregates: providers_count and min prices across available offers.

Param Default Meaning
model_type — text | image | video | speech | transcription | embeddings | rerank | decisions | audio
vendor — exact vendor, e.g. anthropic, moonshotai
q — substring match on model name/slug/vendor
limit · offset 100 · 0 pagination

GET /api/v1/models/{slug}

One model plus every provider offer: prices, upstream model id, effective context window, availability, last check time.

Param Default Meaning
available — true / false — filter offers by availability

GET /api/v1/pricing

Flat provider × model offer matrix — the "who is cheapest" endpoint. Rows join model and provider fields.

Param Default Meaning
sort — price_input | price_output — ascending, offers without a price last
model — substring match on model name/slug
provider — provider slug
model_type — same taxonomy as /models
available true true / false / omit to include everything
limit · offset 100 · 0 pagination 
GET /api/v1/pricing?model=deepseek-v3.2&sort=price_input&limit=1

{
  "items": [{
    "provider_slug": "deepinfra",
    "model_slug": "deepseek-v3-2",
    "price_input_per_mtok": "0.270000",
    "price_output_per_mtok": "0.400000",
    "context_window": 163840,
    "is_available": true,
    "checked_at": "2026-09-26T12:00:00Z"
  }],
  "meta": { "total": 12, "limit": 1, "offset": 0 }
}

GET /api/v1/gpus

Dedicated GPU rental offers, priced per hour.

Param Default Meaning
category — containers | llm_deploy
gpu_type — substring, e.g. B200, H100
provider — provider slug
available — true / false
limit · offset 100 · 0 pagination

GET /api/v1/agent-plans

Managed agent-framework hosting plans across providers.

Param Default Meaning
agent_slug — framework id, e.g. openclaw
provider — provider slug
limit · offset 100 · 0 pagination

GET /api/v1/benchmarks

Curated benchmark scores for catalog models.

Param Default Meaning
benchmark — substring, e.g. MMLU
model — substring match on model name/slug
limit · offset 100 · 0 pagination

Errors

Unknown slug → 404 {"detail": "..."}. Invalid parameter (e.g. limit > 500, unsupported sort) → 422 with a field-level description. Rate limited → 429 plus Retry-After seconds.

Also for machines

  • GET /api/search/suggest?q=claude — JSON search suggestions (the same data that powers the site search box): matching models with the cheapest input price and providers, with relative links into the catalog.
  • GET /llms.txt — a plain-text index of the whole catalog in the llms.txt convention: a short site summary followed by links to every provider and model page.
  • GET /sitemap.xml — a standard XML sitemap of all indexable pages for search engines.