wtt://prices/all
LIVE RATE-CARD REGISTRY
Compare public list prices across frontier labs and inference providers—with context band, delivery mode, cache meters, freshness, and official evidence attached.
Loading verified rate cards…
Standard, Batch, Flex, Priority, and Fast are delivery modes. Cache read and cache write are separate provider-billed meters inside a mode; they apply only to eligible reused or newly cached input.
| Conditions | Rate status | Evidence |
|---|
Try a broader model name or choose all providers.
API QUICKSTART
curl --fail-with-body --get 'https://whatsthetoken.com/v1/prices' \
-H 'Accept: application/json' \
--data-urlencode 'provider=openai' \
--data-urlencode 'mode=standard'The read API requires no authentication and answers cross-origin browser requests. Administration routes and raw archived snapshot bytes remain owner-only.
| Parameter | Accepted value |
|---|---|
provider | Exact provider slug. Required by /v1/price. |
model | Exact provider-scoped API ID or alias. Required by /v1/price. |
mode | standard, batch, flex, priority, or fast. |
region | Exact region label, such as global. |
context_tokens | Non-negative pricing-band token count. Check thresholdRule to see whether the provider counts input, prompt, or context tokens. |
schedule | Exact pricing schedule label. |
customer_tier | Exact published tier, such as public_list. |
service_family | Exact provider offering family, such as inference. |
dimension[<key>] | Exact provider-specific dimension; repeat once per distinct key. |
as_of | ISO 8601 timestamp; defaults to now. |
include_provisional | Boolean; defaults to false. |
/v1/prices lists every applicable variant. /v1/price
requires provider and model and returns 409 when more dimensions are needed.
{
"data": [{
"provider": { "slug": "openai", "display_name": "OpenAI Direct" },
"requested_model": null,
"resolved_model": { "canonical_slug": "gpt-5.6-sol" },
"identifiers": [{ "namespace": "openai.api", "value": "gpt-5.6-sol", "kind": "api_id" }],
"rate_card_id": "…",
"rates": [{ "meter": "input_token", "amount": "…", "currency": "USD", "per": "1000000", "unit": "token" }],
"conditions": { "service_mode": "standard", "region": "global", "customer_tier": "public_list", "dimensions": {} },
"verification_status": "verified"
}],
"meta": { "schema_version": "1.0", "dataset_revision": "…" }
}Amounts and denominators are exact strings. A response object is one rate-card variant—not one model—and meters can be absent when a provider does not publish them.
Checking the latest global refresh…
Prices are public list price per provider-billed token, not normalized task cost. Tokenizers, quality, reasoning tokens, and tool fees differ.