Consommez les données

Référence API

Vous voulez vos données sur TKX ? Vous êtes sur la mauvaise page. Cette page concerne la consommation des données TKX. Si vous exploitez un routeur ou agrégateur (comme 0G) et voulez vos prix dans les classements, vous n’appelez pas cette API : vous exposez un endpoint standard et TKX le récupère automatiquement. → Le standard Open Router · Être référencé

Tous les endpoints sont en GET, renvoient du JSON, ne nécessitent aucune clé et s'actualisent chaque heure. URL de base https://tkx.org. Les exemples ci-dessous sont de vraies réponses, tronquées.

GET/api/v1/rankings.jsonClassements de modèles (top 60)

GET/api/v1/models/{id}.jsonDétail par modèle

GET/api/v1/providers.jsonTableau de confiance des fournisseurs

GET/api/v1/aggregators.jsonTableau des routeurs / agrégateurs

GET/api/v1/indices.jsonIndices TTPI + GPU

GET/api/v1/meta.jsonMétadonnées de collecte

Classements

GET/api/v1/rankings.json

Les 60 meilleurs modèles classés par l'offre vérifiée la moins chère. Chaque entrée porte la meilleure offre, le nombre de fournisseurs, les TTFT/TPS/disponibilité sondés lorsqu'ils sont disponibles, un spark de prix sur 30 jours et le tableau providers complet.

Requête

curl https://tkx.org/api/v1/rankings.json

Réponse (tronqué)

{
  "generatedAt": "2026-07-22T16:07:24.242Z",
  "attribution": "Data by TKX (https://tkx.org) — free use requires attribution",
  "rankings": [
    {
      "id": "gpt-5.5",
      "name": "gpt-5.5",
      "org": "azure_ai",
      "ctx": 1050000,
      "best": {
        "in": 5,
        "out": 30,
        "blended": 11.25,
        "router": "official",
        "provider": "azure_ai",
        "evidence": "official"
      },
      "avgBlended": 11.25,
      "providerCount": 5,
      "ttft": null,
      "tps": null,
      "uptime": null,
      "spark": [],
      "providers": [
        {
          "router": "official",
          "provider": "azure_ai",
          "in": 5,
          "out": 30,
          "ttft": null,
          "tps": null,
          "uptime": null,
          "evidence": "official"
        }
      ],
      "usage": {
        "tokens": 98987458250,
        "tokensPaid": 98987458250,
        "tokensFree": 0,
        "share": 1.03,
        "rank": 15,
        "usd": 1113608.91,
        "volChgPct": null,
        "spendChgPct": null
      }
    }
  ]
}

Détail du modèle

GET/api/v1/models/{id}.json

Chaque offre de fournisseur pour un même modèle canonique, avec les prix et la preuve.

Paramètre de chemin	Type	Description
`id`	string	Identifiant canonique de modèle, ex. `gpt-oss-120b`

Requête

curl https://tkx.org/api/v1/models/gpt-oss-120b.json

Réponse (tronqué)

{
  "generatedAt": "2026-07-06T07:10:53.381Z",
  "attribution": "Data by TKX (https://tkx.org)",
  "model": {
    "id": "claude-3-5-sonnet",
    "name": "claude-3-5-sonnet",
    "org": "snowflake",
    "ctx": 200000,
    "best": {
      "in": 3,
      "out": 15,
      "blended": 6,
      "router": "official",
      "provider": "snowflake",
      "evidence": "official"
    },
    "providerCount": 3,
    "ttft": null,
    "tps": null,
    "uptime": null,
    "spark": [],
    "providers": [
      {
        "router": "official",
        "provider": "snowflake",
        "in": 3,
        "out": 15,
        "ttft": null,
        "tps": null,
        "uptime": null,
        "evidence": "official"
      }
    ]
  }
}

Fournisseurs

GET/api/v1/providers.json

Le tableau de confiance des fournisseurs : couverture, nombre d'offres, nombre de meilleurs prix gagnants, disponibilité/latence médiane (lorsque sondée) et plancher de précision.

Requête

curl https://tkx.org/api/v1/providers.json

Réponse (tronqué)

{
  "attribution": "Data by TKX (https://tkx.org) — free use requires attribution",
  "data": [
    {
      "key": "requesty",
      "name": "requesty",
      "models": 374,
      "offers": 374,
      "bestWins": 306,
      "routers": [
        "requesty"
      ],
      "medUptime": null,
      "upN": 0,
      "medTtft": null,
      "ttftN": 0,
      "medTps": null,
      "evidence": "api",
      "kind": "provider",
      "quant": null,
      "quantN": 0
    }
  ]
}

GET/api/v1/providers/{key}.json

Détail par fournisseur — les modèles qu'un fournisseur sert et comment il se compare.

Paramètre de chemin	Type	Description
`key`	string	Clé de fournisseur, ex. `fireworks`, `0g`

Requête

curl https://tkx.org/api/v1/providers/fireworks.json

Réponse (tronqué)

{
  "generatedAt": "2026-07-22T16:07:24.242Z",
  "attribution": "Data by TKX (https://tkx.org)",
  "provider": {
    "key": "0g",
    "name": "0G Compute",
    "models": 17,
    "offers": 17,
    "bestWins": 14,
    "routers": [
      "0g"
    ],
    "medUptime": null,
    "upN": 0,
    "medTtft": null,
    "ttftN": 0,
    "medTps": null,
    "evidence": "api",
    "kind": "web3",
    "quant": null,
    "quantN": 0,
    "items": [
      {
        "id": "qwen3-vl-30b",
        "name": "qwen3-vl-30b",
        "org": "",
        "in": 0.0194,
        "out": 0.1892,
        "ttft": null,
        "tps": null,
        "uptime": null,
        "evidence": "api",
        "router": "0g",
        "isBest": true
      }
    ]
  }
}

Agrégateurs

GET/api/v1/aggregators.json

Le tableau des routeurs/agrégateurs : markupPct mesuré par rapport au prix affiché, couverture, disponibilité médiane, niveau (live | listed) et région.

Requête

curl https://tkx.org/api/v1/aggregators.json

Réponse (tronqué)

{
  "attribution": "Data by TKX (https://tkx.org) — free use requires attribution",
  "data": [
    {
      "slug": "openrouter",
      "name": "OpenRouter",
      "kind": "aggregator",
      "url": "https://openrouter.ai",
      "models": 319,
      "offers": 612,
      "markupPct": 0,
      "markupN": 199,
      "medUptime": 99.5547,
      "medTtft": null,
      "evidence": "api",
      "tier": "live",
      "region": "western",
      "note": "",
      "oai": true,
      "trustScore": 90,
      "completeness": 100,
      "breakdown": {
        "coverage": 100,
        "price": 80,
        "reliability": 93,
        "evidence": 80
      },
      "volume24h": {
        "tokens": 9621566458152,
        "usd": 6402826,
        "date": "2026-07-17",
        "stale": true,
        "volChgPct": null,
        "spendChgPct": null
      }
    }
  ]
}

Indices

GET/api/v1/indices.json

Indice du prix des tokens TTPI plus indices spot GPU (international en direct + domestique curé), chacun avec la taille d'échantillon et la version de méthodologie.

Requête

curl https://tkx.org/api/v1/indices.json

Réponse (tronqué)

{
  "generatedAt": "2026-07-22T16:07:24.241Z",
  "methodology": {
    "version": "M-2026.1",
    "url": "https://tkx.org/methodology/"
  },
  "indices": {
    "TTPI": {
      "name": "TKX 词元价格指数",
      "unit": "USD / 1M tokens (completion, median)",
      "value": 1.9,
      "sample": 319
    },
    "TTV-OR": {
      "name": "TKX Token Traded Value (OpenRouter, est.)",
      "unit": "USD / day (est., paid traffic at list rates, 3:1 in:out blend)",
      "value": 6402826,
      "sample": 36,
      "date": "2026-07-17",
      "stale": true,
      "note": "Estimated list-rate value of paid token traffic routed through OpenRouter. Not billed revenue; free-tier traffic excl…"
    }
  },
  "gpu": {
    "international": [
      {
        "gpu": "H200",
        "source": "vast",
        "currency": "USD",
        "priceHr": 4.08,
        "sample": 18,
        "history": []
      }
    ],
    "domestic": [
      {
        "gpu": "RTX 4090",
        "source": "curated snapshot 2026-07-04",
        "currency": "CNY",
        "priceHr": 2.18,
        "sample": 5,
        "history": []
      }
    ]
  },
  "attribution": "Data by TKX (https://tkx.org) — free use requires attribution"
}

Variations

GET/api/v1/changes.json

Événements de variation de prix sur 24 heures — les mouvements du prix mixte le moins cher par modèle au-dessus du seuil, d'un jour à l'autre.

Requête

curl https://tkx.org/api/v1/changes.json

Réponse (tronqué)

{
  "attribution": "Data by TKX (https://tkx.org) — free use requires attribution",
  "generatedAt": "2026-07-22T16:07:24.211Z",
  "window": "24h",
  "events": []
}

GET/api/v1/trending.json

Principales évolutions sur 30 jours et applications tendance.

Requête

curl https://tkx.org/api/v1/trending.json

Réponse (tronqué)

{
  "attribution": "Data by TKX (https://tkx.org) — free use requires attribution",
  "generatedAt": "2026-07-22T16:07:24.211Z",
  "topMovers": [],
  "topApps": [
    {
      "title": "Command Code",
      "url": "https://commandcode.ai/",
      "growth": 1431
    }
  ]
}

Applications

GET/api/v1/apps.json

Tableau d'usage des tokens par applications et agents, attribué à OpenRouter.

Requête

curl https://tkx.org/api/v1/apps.json

Réponse (tronqué)

{
  "attribution": "App/agent token usage via OpenRouter (openrouter.ai/apps)",
  "generatedAt": "2026-07-22T16:07:24.203Z",
  "source": "OpenRouter",
  "popular": [
    {
      "rank": 2,
      "title": "Hermes Agent",
      "desc": "Hermes Agent is an open-source, self-improving AI agent by Nous Research that runs persistently with memory across se…",
      "slug": "hermes-agent",
      "url": "https://hermes-agent.nousresearch.com/",
      "cats": [
        "personal-agent"
      ],
      "tokens": 28772524995746,
      "requests": 367166827
    }
  ],
  "trending": [
    {
      "rank": 7,
      "title": "Command Code",
      "desc": "",
      "slug": "",
      "url": "https://commandcode.ai/",
      "cats": [
        "programming-app"
      ],
      "tokens": 1273389568214,
      "requests": 37968056,
      "growth": 1431
    }
  ],
  "leaderboards": {
    "coding": [
      {
        "rank": 2,
        "title": "Hermes Agent",
        "desc": "Hermes Agent is an open-source, self-improving AI agent by Nous Research that runs persistently with memory across se…",
        "slug": "hermes-agent",
        "url": "https://hermes-agent.nousresearch.com/",
        "cats": [
          "personal-agent"
        ],
        "tokens": 889819065274,
        "requests": 12846144
      }
    ],
    "creative": [
      {
        "rank": 9,
        "title": "Descript",
        "desc": "AI Video & Podcast Editor",
        "slug": "descript",
        "url": "https://descript.com/",
        "cats": [
          "video-gen"
        ],
        "tokens": 72758705917,
        "requests": 919852
      }
    ],
    "productivity": [
      {
        "rank": 2,
        "title": "Hermes Agent",
        "desc": "Hermes Agent is an open-source, self-improving AI agent by Nous Research that runs persistently with memory across se…",
        "slug": "hermes-agent",
        "url": "https://hermes-agent.nousresearch.com/",
        "cats": [
          "personal-agent"
        ],
        "tokens": 889819065274,
        "requests": 12846144
      }
    ],
    "entertainment": [
      {
        "rank": 12,
        "title": "Hello Minds, powered by Ethoswarm",
        "desc": "",
        "slug": "",
        "url": "https://www.hellominds.ai/",
        "cats": [
          "roleplay"
        ],
        "tokens": 50652424750,
        "requests": 233112
      }
    ]
  }
}

Clouds GPU

GET/api/v1/gpu-clouds.json

Tableau des prix des fournisseurs de cloud GPU — instantané curé plus prix spot Vast en direct, par carte.

Requête

curl https://tkx.org/api/v1/gpu-clouds.json

Réponse (tronqué)

{
  "attribution": "Data by TKX (https://tkx.org) — free use requires attribution",
  "asof": "2026-07-04",
  "vastLiveAt": "2026-07-22T11:39:11.061Z",
  "clouds": [
    {
      "name": "Vast.ai",
      "url": "https://vast.ai/",
      "region": "spot",
      "cur": "$",
      "p": {
        "4090": 0.369,
        "5090": 0.457,
        "a100": 0.889,
        "h100": 2.235,
        "b200": 7.126
      },
      "tier": "live",
      "note": "竞价市场,最便宜;会被抢占需 checkpoint。TKX 实时抓取此源。"
    }
  ]
}

Historique

GET/api/v1/history/ttpi.json

Historique quotidien du TTPI (fenêtre de 30 jours).

Requête

curl https://tkx.org/api/v1/history/ttpi.json

Réponse (tronqué)

{
  "attribution": "Data by TKX (https://tkx.org) — free use requires attribution",
  "code": "TTPI",
  "unit": "USD per 1M completion tokens (median)",
  "history": []
}

GET/api/v1/history/gpu.json

Historique quotidien des prix GPU par source et par carte (fenêtre de 30 jours).

Requête

curl https://tkx.org/api/v1/history/gpu.json

Réponse (tronqué)

{
  "attribution": "Data by TKX (https://tkx.org) — free use requires attribution",
  "unit": "per GPU-hour (vast=USD live market, curated-cn=CNY curated snapshot)",
  "series": {}
}

Méta

GET/api/v1/meta.json

Métadonnées de collecte — décomptes et statut par source (y compris les sources auxiliaires dev/apps/gpu).

Requête

curl https://tkx.org/api/v1/meta.json

Réponse (tronqué)

{
  "generatedAt": "2026-07-22T16:07:24.242Z",
  "counts": {
    "models": 1593,
    "offers": 3177,
    "pages": 207
  },
  "sources": [
    {
      "source": "litellm",
      "ok": true,
      "count": 2068
    }
  ]
}

Spécification OpenAPI

GET/api/v1/openapi.json

Le contrat complet lisible par machine (OpenAPI 3.1). Pointez votre codegen / Swagger UI dessus.

Requête

curl https://tkx.org/api/v1/openapi.json

Réponse (tronqué)

{
  "openapi": "3.1.0",
  "info": {
    "title": "TKX Public Data API",
    "version": "1.3.0",
    "description": "The data authority for the AI token & compute market. Free with attribution to tkx.org. All endpoints are static JSON…",
    "contact": {
      "email": "echo@minax.io",
      "url": "https://tkx.org/standard/"
    }
  },
  "servers": [
    {
      "url": "https://tkx.org"
    }
  ],
  "paths": {
    "/api/v1/rankings.json": {
      "get": {
        "summary": "Model rankings: sorted by est. 24h traded value first (usage.usd = paid tokens × list rate; free traffic never priced…",
        "responses": {
          "200": {
            "description": "OK"
          }
        }
      }
    },
    "/api/v1/models/{id}.json": {
      "get": {
        "summary": "Per-model detail: all provider offers, prices, evidence",
        "parameters": [
          {
            "name": "id",
            "in": "path",
            "required": true,
            "schema": {
              "type": "string"
            }
          }
        ],
        "responses": {
          "200": {
            "description": "OK"
          },
          "404": {
            "description": "Unknown model"
          }
        }
      }
    },
    "/api/v1/providers.json": {
      "get": {
        "summary": "Provider trust board: coverage, cheapest-wins, uptime/latency medians, precision floor",
        "responses": {
          "200": {
            "description": "OK"
          }
        }
      }
    },
    "/api/v1/providers/{key}.json": {
      "get": {
        "summary": "Per-provider detail",
        "parameters": [
          {
            "name": "key",
            "in": "path",
            "required": true,
            "schema": {
              "type": "string"
            }
          }
        ],
        "responses": {
          "200": {
            "description": "OK"
          }
        }
      }
    },
    "/api/v1/aggregators.json": {
      "get": {
        "summary": "Router board (45 tracked), sorted by 24h token volume first (Deng 2026-07-21: volume is the primary ranking; coverage…",
        "responses": {
          "200": {
            "description": "OK"
          }
        }
      }
    },
    "/api/v1/indices.json": {
      "get": {
        "summary": "TTPI token price index, TTV-OR est. daily token traded value (paid traffic at list rates), GPU spot indices (intl liv…",
        "responses": {
          "200": {
            "description": "OK"
          }
        }
      }
    },
    "/api/v1/changes.json": {
      "get": {
        "summary": "24h price-change events (>2% day-over-day per-model cheapest blended)",
        "responses": {
          "200": {
            "description": "OK"
          }
        }
      }
    },
    "/api/v1/trending.json": {
      "get": {
        "summary": "Top movers (30d) and trending apps",
        "responses": {
          "200": {
            "description": "OK"
          }
        }
      }
    },
    "/api/v1/usage.json": {
      "get": {
        "summary": "Real per-model token usage/share + est. spend (OpenRouter rankings-daily; paid/free split; est. at list rates; OpenRo…",
        "responses": {
          "200": {
            "description": "OK"
          }
        }
      }
    },
    "/api/v1/apps.json": {
      "get": {
        "summary": "Apps & agents token-usage board (attributed to OpenRouter)",
        "responses": {
          "200": {
            "description": "OK"
          }
        }
      }
    },
    "/api/v1/gpu-clouds.json": {
      "get": {
        "summary": "GPU cloud supplier price table (curated + Vast live)",
        "responses": {
          "200": {
            "description": "OK"
          }
        }
      }
    },
    "/api/v1/history/ttpi.json": {
      "get": {
        "summary": "TTPI daily history (30d window)",
        "responses": {
          "200": {
            "description": "OK"
          }
        }
      }
    },
    "/api/v1/history/gpu.json": {
      "get": {
        "summary": "GPU price daily history by source|card (30d window)",
        "responses": {
          "200": {
            "description": "OK"
          }
        }
      }
    },
    "/api/v1/constituents/ttpi.json": {
      "get": {
        "summary": "TTPI full constituents for today's point + sha256 — recompute/audit the index without database access",
        "responses": {
          "200": {
            "description": "OK"
          }
        }
      }
    },
    "/api/v1/constituents/ttv-or.json": {
      "get": {
        "summary": "TTV-OR full constituents (per-model est. spend) + sha256 for today's point",
        "responses": {
          "200": {
            "description": "OK"
          }
        }
      }
    },
    "/api/v1/price-ohlc.json": {
      "get": {
        "summary": "Per-model daily price OHLC (cheapest blended, 90d) — real prices, candles gain intraday range as hourly buckets accum…",
        "responses": {
          "200": {
            "description": "OK"
          }
        }
      }
    },
    "/api/v1/meta.json": {
      "get": {
        "summary": "Collection metadata: counts, source status incl. aux (dev/apps/gpu)",
        "responses": {
          "200": {
            "description": "OK"
          }
        }
      }
    }
  }
}

← Précédent

Attribution et limites

Modèle de données et preuve

Référence API

Classements

Détail du modèle

Fournisseurs

Agrégateurs

Indices

Variations

Tendances

Applications

Clouds GPU

Historique

Méta

Spécification OpenAPI