⬡Inference Index
API & Data

Programmatic access

Every page on this site is backed by a public JSON feed. No API key, no rate limiting for reasonable use, open CORS. Build on top of it — agents, dashboards, research papers.

GET /api/prices

Live enriched pricing — every tracked model with all providers, best price flagged, and CheapTokens live discount applied. Cached 30s at the edge.

$ curl https://inferenceindex.ai/api/prices
{
  "models": [
    {
      "id": "claude-opus-4-6",
      "name": "Claude Opus 4.6 (Beta)",
      "enrichedProviders": [
        { "name": "CheapTokens", "input": 3.47, "output": 17.36, "isBest": true },
        { "name": "Anthropic",   "input": 5.50, "output": 27.50 }
      ],
      "cheapest": { "provider": "CheapTokens", "input": 3.47, "output": 17.36, "savings": "37%" }
    }
  ],
  "stats": { "modelCount": 69, "providerCount": 9 }
}

GET /api/snapshot

Daily snapshot of the index — II-C / II-I / II-V values, frontier model, top-10, price distribution. Query ?date=YYYY-MM-DD for a specific day, or omit for latest.

$ curl https://inferenceindex.ai/api/snapshot
$ curl https://inferenceindex.ai/api/snapshot?date=2026-04-16
{
  "date": "2026-04-16",
  "inputIndex": 1.421,
  "outputIndex": 7.563,
  "blendedIndex": 2.838,
  "frontierModel": "gemini-3-1-pro-preview",
  "frontierModelName": "Gemini 3.1 Pro Preview",
  "frontierIQ": 57.2,
  "intelligenceCeiling": 57.2,
  "avgIQ": 47.49,
  "top10": [ ... 10 entries ... ]
}

GET /api/intelligence

Intelligence scores (Artificial Analysis Intelligence Index), coding, math, speed, and TTFT for every tracked model. Cached 6h in-memory with a static fallback.

$ curl https://inferenceindex.ai/api/intelligence
{
  "models": {
    "claude-opus-4-6": {
      "score": 53.0,
      "coding": 60.0,
      "math": 54.0,
      "speedTps": 68,
      "ttftSeconds": 2.1,
      "blendedPrice": 8.94
    }
  },
  "source": "artificial-analysis",
  "cachedAt": "2026-04-16T16:00:00Z"
}

GET /api/venice-models

Raw live model list — base Venice prices plus our manually-verified direct provider overrides (Anthropic, OpenAI, Google, xAI, DeepSeek, Groq, Together). Cached 5min.

$ curl https://inferenceindex.ai/api/venice-models

Machine-readable descriptions

For AI agents crawling this site or using it as a tool:

For journalists & researchers

Cite any index value with a date. The snapshot for that date is immutable once written — fetchable at /api/snapshot?date=YYYY-MM-DD.

Suggested citation format:

Inference Index (v1.1). "II-C Cost Index, {date}." https://inferenceindex.ai/api/snapshot?date={date}

Example: agent tool-use

Register the Inference Index as a tool your agent can call when it needs current pricing context before choosing a model. The JSON is small (<50KB), refreshed every 30s, and open-CORS.

const res = await fetch('https://inferenceindex.ai/api/prices');
const { models } = await res.json();
const candidate = models
  .filter(m => m.cheapest)
  .sort((a, b) => a.cheapest.input - b.cheapest.input)[0];
console.log(`Cheapest inference right now: ${candidate.name} @ $${candidate.cheapest.input}/1M`);

Questions, bugs, or coverage requests? Open an issue or email hello@inferenceindex.ai. See methodology.