Quote Check
Evaluate a vendor quote in memory against historical published list prices, peer spreads, and reference rates.
POST /api/v1/quote-check
GET /api/v1/quote-check# Evaluate in memory with POST
curl -X POST https://vitna.ai/api/v1/quote-check \
-H "Content-Type: application/json" \
-d '{
"sku": "together-kimi-k3",
"quotedAmountUsd": 2.50,
"dimension": "input_per_mtok"
}'
# Or check via query parameters
curl "https://vitna.ai/api/v1/quote-check?sku=lambda-h100-sxm&amount=2.49&dimension=gpu_hour_on_demand"
# A long-context quote on a model priced in context bands
curl "https://vitna.ai/api/v1/quote-check?sku=gpt-5.4-pro&amount=60&dimension=input_per_mtok&promptTokens=300000"Zero retention guarantee
A pasted quote is confidential enterprise procurement data. The request is evaluated purely in memory and is never written to disk, database, or server logs. Responses carry Cache-Control: no-store.
Parameters
| Field | Type | Description |
|---|---|---|
sku | string | SKU slug or ID (e.g. together-kimi-k3, gpt-5.6-sol, lambda-h100-sxm). Required. |
quotedAmountUsd | number | Quoted price in USD per unit. Required. |
dimension | string | Meter dimension (e.g. input_per_mtok, gpu_hour_on_demand). Optional; defaults to class standard. |
date | string | ISO date YYYY-MM-DD. Optional; defaults to today. Future dates are refused. |
promptTokens | integer | Prompt length, in tokens, of the requests the quote covers. Optional. On a model priced in context bands it picks the band the quote is compared with; without it, the band from the first token. A whole number, 0 or more. As a query parameter it is digits only, so 300,000 is refused rather than read as 300. |
Response
The answer to the first request above, as the record stood on 17 November 2026:
{
"quote": {
"quotedAmountUsd": 2.5,
"asOfDate": "2026-11-17",
"unit": "USD per 1M tokens",
"promptTokens": null
},
"target": {
"skuId": "together/kimi-k3",
"skuSlug": "together-kimi-k3",
"skuName": "Kimi K3",
"providerId": "together",
"classId": "model",
"product": "Kimi K3",
"productSlug": "kimi-k3",
"dimensionId": "input_per_mtok",
"dimensionLabel": "Input price",
"unit": "USD per 1M tokens"
},
"vendorList": {
"found": true,
"source": "database",
"listPriceUsd": 3,
"unit": "USD per 1M tokens",
"bandFromTokens": 0,
"bandToTokens": null,
"bandLabel": null,
"validFrom": "2026-10-12T07:08:29.430Z",
"validTo": null,
"confirmed": true,
"confirmedAt": "2026-11-17T17:10:09.011Z",
"evidenceSha256": "a0843851a01bb156386380c57a7c21a7e4a0b35845c3e1e6c4dcafa0d36702e7",
"sourceUrl": "https://www.together.ai/pricing",
"bands": [
{
"bandFromTokens": 0,
"bandToTokens": null,
"bandLabel": null,
"listPriceUsd": 3
}
],
"discountVsListPct": 16.7,
"listComparisonKind": "discount"
},
"peerSpread": {
"available": true,
"product": "Kimi K3",
"productSlug": "kimi-k3",
"dimension": "input_per_mtok",
"unit": "USD per 1M tokens",
"providerCount": 6,
"minUsd": 2.85,
"maxUsd": 3.3,
"spreadPct": 16,
"position": "below_cheapest",
"quotedVsPeerMinPct": 12.3,
"quotedVsPeerMaxPct": null,
"providers": [
{
"provider": "DeepInfra",
"skuSlug": "deepinfra-kimi-k3",
"amountUsd": 2.85,
"derived": false,
"stale": false
},
{
"provider": "Parasail",
"skuSlug": "parasail-kimi-k3",
"amountUsd": 3,
"derived": false,
"stale": false
},
{
"provider": "Together AI",
"skuSlug": "together-kimi-k3",
"amountUsd": 3,
"derived": false,
"stale": false
},
{
"provider": "Novita AI",
"skuSlug": "novita-kimi-k3",
"amountUsd": 3,
"derived": false,
"stale": false
},
{
"provider": "Baseten",
"skuSlug": "baseten-kimi-k3",
"amountUsd": 3,
"derived": false,
"stale": false
},
{
"provider": "AWS",
"skuSlug": "aws-kimi-k3",
"amountUsd": 3.3,
"derived": true,
"stale": false
}
],
"note": null
},
"referenceRate": {
"available": true,
"medianUsd": 3,
"minUsd": 2.85,
"maxUsd": 3.3,
"minProviders": 3,
"providerCount": 6,
"method": "median-of-confirmed-list-v1",
"coversOn": "2026-11-16",
"discountVsReferenceRatePct": 16.7,
"note": null
},
"notices": {
"scope": "Published list price comparison only. Enterprise quotes reflect bilateral commitments, volume tiers, support SLAs, and geographical routing terms that public list prices do not capture.",
"privacy": "Quotes are evaluated in memory and never stored. Bilateral enterprise terms remain confidential.",
"band": null
},
"meta": {
"generatedAt": "2026-11-17T17:28:04.597Z",
"source": "database"
}
}A peerSpread on a cache-write meter also carries blockNote, the same qualification a spread carries: sellers cache in blocks of their own size, so equal rates need not buy the same caching.
Context bands
Some models are priced by prompt length, one list price up to a threshold and another above it, and the record keeps a price per band. A quote is for requests of some length, so it can say which.
- `promptTokens` picks the band. The quote is compared with the band that holds a prompt of that many tokens. The record's lower bound is exclusive and its upper bound inclusive, so a prompt of exactly 272,000 tokens is billed in the band to 272K tokens.
- Without it, the band from the first token is compared. That is the rate at the shortest prompt length the provider prices, and on a banded model it is not what a long prompt pays.
notices.bandsays so in words. - The answer names the band either way.
vendorList.bandFromTokens,bandToTokensandbandLabeldescribe the band compared, andvendorList.bandslists every band the meter is priced in on that date, with its list price, from the shortest prompt up. A meter priced once has one band,(0, null], with a null label, andnotices.bandis null. - A prompt no band holds has no list price. Some meters are priced only up to a threshold. The list check then answers
found: false, its band keys null, with the bands it does have listed.
Peer spreads and reference rates are drawn over each seller's rate at the shortest prompt length it prices, which is the only band every seller on a peer set publishes (spreads). So a quote in a band above the first token is compared with neither: peerSpread answers available: false with position: "not_compared", its counts and figures null and a note saying why, and referenceRate answers available: false with a note. A seller priced once has a single band from the first token, so its quote is compared at any prompt length, against the other sellers' shortest-band rates, which is what a spread holds.
{
"quote": { "quotedAmountUsd": 60, "asOfDate": "2026-11-16", "unit": "USD per 1M tokens", "promptTokens": 300000 },
"vendorList": {
"found": true,
"listPriceUsd": 60,
"bandFromTokens": 272000,
"bandToTokens": null,
"bandLabel": "over 272K tokens",
"bands": [
{ "bandFromTokens": 0, "bandToTokens": 272000, "bandLabel": "to 272K tokens", "listPriceUsd": 30 },
{ "bandFromTokens": 272000, "bandToTokens": null, "bandLabel": "over 272K tokens", "listPriceUsd": 60 }
],
"discountVsListPct": 0,
"listComparisonKind": "at_list"
},
"peerSpread": {
"available": false,
"providerCount": null,
"position": "not_compared",
"note": "Peer spreads compare each seller's rate at the shortest prompt length it prices, and this quote is in the band over 272K tokens, so none is compared."
},
"notices": {
"band": "The record prices this meter in 2 context bands: to 272K tokens and over 272K tokens. A prompt of 300,000 tokens is billed in the band over 272K tokens, and the quote is compared with that band's list price."
}
}The same quote with no promptTokens is compared with the $30 band and reads as a premium, and its notices.band says that it was compared with the band to 272K tokens and that sending promptTokens compares it with the band the prompt is billed in. That default is what this check always did, and until 2026-11-16 it did it without saying so.
Published list boundary
Enterprise quotes carry volume commitments, dedicated capacity, latency SLAs, and custom routing terms that public list prices do not include. Quote Check compares strictly against the vendor published list on that date.
