API | readable AI benchmarks

Overview

This page documents the json routes. The Graph on / is a scatter of Blended price vs Quality Score / Value Score. It is not the ranking. Ranked pick is GET /api/recommend. No API key.

models[] is the requested ranking (default: Quality Score)
price[] is Value Score (quality per dollar). It is not cheapest Blended price.
cheap[] is lowest Blended price above 0
lookup: ?q=NAME (name / slug / provider)

Rankings

Field	Type	Description
Quality Score	number 0-100	`precision = Correct / (Correct + Incorrect); excessWrong = max(0, Incorrect − Correct); answerQuality = clamp((0.6 × Correct + 0.4 × precision) × (1 − excessWrong × 0.75)); Quality Score = (Σ available weight × component / Σ available weight) × 100, with weights 0.50 / 0.20 / 0.075 / 0.075 on answerQuality, clamp(Intelligence/65), clamp(Coding/80), clamp(Agentic/60); missing index drops from both sums` Formula v1; answerQuality has a 0.50 pre-normalization weight and all available weights normalize. Published AA indices are combined without source-overlap correction; higher is better.
Value Score	number 0-100	`capability = clamp(Intelligence / 65); reliability = clamp(Correct × (1 − Incorrect × 0.35)) with missing Incorrect = 0; affordability = 1 / (1 + log10(1 + max(0, blendedPrice) × 8) × 0.45); Value Score = (0.6 × capability + 0.4 × reliability) × affordability × 100` Formula v1; always uses blendedPrice, not selected graph cost. higher is better.
Reliability	number 0-100	`Reliability = clamp((correct − incorrect + 1) / 2, 0, 1) × 100; requires finite correct and incorrect; abstain is not an input` Formula v1; Graph Y-axis only. higher is better.
Blended price	USD / 1M tokens	7:2:1 cache-hit · input · output. lower is better.
Intelligence Index	number	live data Intelligence Index on the Graph and Shortlist.
Coding Index	number	live data Coding Index.
Agentic Index	number	live data Agentic Index.
Correct answers	rate 0-1	share of Omniscience questions answered correctly.
Incorrect answers	rate 0-1	share of Omniscience questions answered incorrectly.

cheap[] skips blendedPrice of 0 (local / unpublished rows).

Authentication

None. GET, HEAD, OPTIONS. CORS *.

GET`/api/recommend`

ranked pick for a constraint set. Default goal=quality, limit=10. Default body also includes price[] and cheap[]. read is the first field after source.

Query

Key	Type	Default	Description
`goal`	`quality` \| `price` \| `cheap` \| `coding` \| `intelligence` \| `evidence`	`quality`	quality: Quality Score. price: Value Score (not cheapest USD). cheap: lowest Blended price above 0. coding / intelligence: those indexes. evidence: published evidence score.
`maxBlendedPrice`	finite number	none	USD / 1M Blended price ceiling.
`minIntelligence`	finite number	none	minimum Intelligence Index.
`minQualityValue`	finite number	none	minimum Quality Score.
`reasoning`	`true` \| `false`	none	keep only reasoning or non-reasoning rows.
`openWeights`	`true` \| `false`	none	open or closed weights.
`minContextWindow`	integer tokens	none	minimum context window.
`inputModality`	comma list `text,image,video,speech`	none	AND: model must include every listed input.
`outputModality`	comma list `text,image,video,speech`	none	AND: model must include every listed output.
`provider`	string, max 64	none	case-insensitive provider substring.
`q`	string, max 64	none	lookup by name / slug / provider. example: ?q=grok
`limit`	integer 1-10	`10`	row cap. 11 is 400.

Unknown keys return 400.

Response fields

Field	Type	Description
`source`	string	always `recommend`
`read`	string	first field after source. models[] is the requested ranking. price[] is Value Score, not cheapest USD. cheap[] is lowest Blended price. lookup: ?q=NAME.
`goal`	string	applied goal.
`updatedAt`	ISO datetime	from the models feed, else now.
`note`	string	live ranking. HTML is a graph view, not the ranking. do not scrape /.
`filters`	object	echo of applied query keys.
`count`	integer	length of models[].
`models`	array	requested ranking.
`price`	array	Value Score ranking. omitted when goal=price.
`cheap`	array	lowest Blended price above 0. omitted when goal=cheap.

models[] row

Field	Type	Description
`slug`	string	canonical slug.
`name`	string	display name.
`provider`	string	creator name.
`reasoning`	boolean or null	reasoning row, if listed.
`openWeights`	boolean or null	open weights, if listed.
`contextWindowTokens`	integer or null	context window.
`inputModalities`	string[]	text / image / video / speech.
`outputModalities`	string[]	text / image / video / speech.
`blendedPrice`	number or null	Blended price, USD / 1M.
`intelligence`	number or null	Intelligence Index.
`coding`	number or null	Coding Index.
`agentic`	number or null	Agentic Index.
`correct`	number or null	Correct answers rate.
`incorrect`	number or null	Incorrect answers rate.
`qualityValue`	number or null	Quality Score 0-100.
`priceValue`	number or null	Value Score 0-100.
`why`	string	one short sentence from actual numbers.

Example Request

curl -sS "https://models.deggo.fyi/api/recommend"
curl -sS "https://models.deggo.fyi/api/recommend?q=grok"
curl -sS "https://models.deggo.fyi/api/recommend?goal=cheap"

from urllib.request import Request, urlopen

req = Request(
    "https://models.deggo.fyi/api/recommend",
    headers={"Accept": "application/json"},
)
with urlopen(req) as response:
    print(response.read().decode())

Example Response

{
  "source": "recommend",
  "read": "models[] is the requested ranking. default goal=quality. price[] is quality per dollar, not cheapest USD. cheap[] is lowest blendedPrice above 0. lookup one model: GET /api/recommend?q=NAME. HTML is a graph view, not the ranking.",
  "goal": "quality",
  "updatedAt": "<from models feed, else ISO now>",
  "note": "live ranking. HTML is a graph view, not the ranking. do not scrape /.",
  "filters": { "goal": "quality", "limit": 10 },
  "count": 10,
  "models": [
    {
      "slug": "",
      "name": "",
      "provider": "",
      "reasoning": true,
      "openWeights": false,
      "contextWindowTokens": 0,
      "inputModalities": [],
      "outputModalities": [],
      "blendedPrice": 0,
      "intelligence": 0,
      "coding": 0,
      "agentic": 0,
      "correct": 0,
      "incorrect": 0,
      "qualityValue": 0,
      "priceValue": 0,
      "why": "short sentence from actual numbers"
    }
  ],
  "price": [],
  "cheap": []
}

Errors

Status	Body	Description
400	`{ "error": "unknown query key: ..." }` `{ "error": "invalid limit" }` `{ "error": "query too long" }`	unknown key, out of range, or query string over 512.
429	`{ "error": "rate limit" }`	Retry-After: 60. 30 / 60s per IP. 120 / 60s global.
503	`{ "error": "models unavailable" }`	catalog loader failed.
405	`{ "error": "method not allowed" }`	not GET, HEAD, or OPTIONS.

GET`/api`

Machine index. Fetch this first if query keys are unknown. Field use explains models[] vs price[] vs cheap[].

Example Request

curl -sS "https://models.deggo.fyi/api"

Other routes

/api/models full catalog the Graph uses. not a ranking. do not hammer. prefer recommend.
/api/benchmarks external boards
/api/subscriptions provider plans
/llms.txt crawler pointer

Limits

30 / 60s per IP on /api and /api/recommend
120 / 60s global
query string <= 512 characters
limit 1-10
no API key
GET / HEAD / OPTIONS only

readable AI benchmarks

Overview

Rankings

Authentication

GET/api/recommend

Query

Response fields

models[] row

Example Request

Example Response

Errors

GET/api

Example Request

Other routes

Limits

GET`/api/recommend`

GET`/api`