Beta version: *Information might not be fully accurate. Please report any discrepancies.

API Reference

API Documentation

REST API for accessing models, benchmarks, scores, and leaderboard data programmatically.

Models

1573

Benchmarks

206

Client-Side Filtering

Since the API is static, filtering must be done client-side. Here are examples:

// Fetch all models
const response = await fetch('/api/v1/models');
const data = await response.json();

// Filter by provider
const openaiModels = data.models.filter(m => m.provider === 'OpenAI');

// Filter by family
const llamaModels = data.models.filter(m => m.family === 'llama');

// Filter by capability
const reasoningModels = data.models.filter(m => 
  m.apiSupport?.reasoning === true
);

// Filter by open source
const openSourceModels = data.models.filter(m => m.isOpenSource);

// Combine filters
const openaiReasoningModels = data.models.filter(m => 
  m.provider === 'OpenAI' && m.apiSupport?.reasoning
);

Base URL

https://llm-registry.com/api/v1

Format

JSON

Auth

None

Endpoints

GET/api/v1

API root with endpoint listing

Sample Response

{
  "apiVersion": "v1",
  "endpoints": { ... },
  "attribution": { ... }
}

GET/api/v1/meta

Registry metadata including counts, categories, and latest score date

Sample Response

{
  "apiVersion": "v1",
  "generatedAt": "2026-02-19T...",
  "latestScoreDate": "2026-02-16",
  "counts": {
    "models": <dynamic>,
    "benchmarks": <dynamic>,
    "categories": <dynamic>,
    "sources": <dynamic>
  },
  "categories": [...],
  "endpoints": [...]
}

GET/api/v1/models

List all models (static pre-generated response)

Query Parameters

limitResults per page (1-500, default: 100) - Client-side only

offsetPagination offset (default: 0) - Client-side only

Sample Response

{
  "total": 150,
  "offset": 0,
  "limit": 100,
  "models": [
    {
      "id": "claude-3-5-sonnet-20241022",
      "name": "Claude 3.5 Sonnet",
      "provider": "Anthropic",
      "family": "claude-sonnet",
      "status": "active",
      "releaseDate": "2024-10-22",
      "trainingCutoff": "2024-04",
      "capabilities": ["text", "vision", "tools"],
      "isOpenSource": false,
      "specs": {
        "contextWindow": 200000,
        "maxOutputTokens": 64000,
        "pricing": {
          "input": 3.0,
          "output": 15.0,
          "cacheInput": 0.3,
          "cacheOutput": 3.75
        }
      },
      "apiSupport": {
        "reasoning": false,
        "vision": true,
        "tools": true,
        "structuredOutput": true,
        "attachment": true
      },
      "coverage": 85.2
    }
  ]
}

GET/api/v1/models/[id]

Get a single model by ID with full details and scores

Sample Response

{
  "model": {
    "id": "claude-3-5-sonnet",
    "name": "Claude 3.5 Sonnet",
    "provider": "Anthropic",
    "scores": {
      "mmlu": { "score": 88.7, "sourceId": "anthropic", ... },
      ...
    }
  }
}

GET/api/v1/benchmarks

List all benchmarks with optional category filter

Query Parameters

categoryFilter by category (e.g., 'Coding', 'Reasoning')

Sample Response

{
  "total": <dynamic>,
  "categories": ["Coding", "Math", "Reasoning", ...],
  "benchmarks": [
    {
      "id": "mmlu",
      "name": "MMLU",
      "category": "Knowledge",
      "maxScore": 100,
      "normalizeMethod": "max"
    }
  ]
}

GET/api/v1/scores

Query scores with flexible filtering

Query Parameters

modelIdFilter by model ID

benchmarkIdFilter by benchmark ID

categoryFilter by benchmark category

sourceIdFilter by data source

limitResults per page (1-5000, default: 500)

offsetPagination offset (default: 0)

Sample Response

{
  "total": <dynamic>,
  "scores": [
    {
      "modelId": "gpt-4o",
      "modelName": "GPT-4o",
      "benchmarkId": "mmlu",
      "benchmarkName": "MMLU",
      "category": "Knowledge",
      "score": 88.7,
      "normalizedScore": 88.7,
      "verified": true,
      "verificationLevel": "third_party",
      "sourceId": "openai",
      "asOfDate": "2024-05-13"
    }
  ]
}

GET/api/v1/leaderboards/[category]

Get ranked leaderboard for a category (use 'all' for global)

Query Parameters

limitMax results (1-500, default: 100)

Sample Response

{
  "category": "Coding",
  "categorySlug": "coding",
  "benchmarkCount": 24,
  "leaderboard": [
    {
      "rank": 1,
      "modelId": "claude-3-5-sonnet",
      "modelName": "Claude 3.5 Sonnet",
      "provider": "Anthropic",
      "average": 89.3,
      "coverage": 95.8,
      "scoreCount": 23
    }
  ]
}

GET/api/v1/export

Export all scores in JSON or CSV format for research workflows

Query Parameters

formatOutput format: 'json' (default) or 'csv'

modelIdFilter by model ID

benchmarkIdFilter by benchmark ID

categoryFilter by benchmark category

sourceIdFilter by data source

Sample Response

{
  "total": <dynamic>,
  "exportedAt": "2026-02-19T...",
  "filters": { ... },
  "scores": [ ... ],
  "attribution": { ... }
}

Static Slicing (Per-Model Files)

For better performance, individual model metadata is available as pre-generated JSON files:

Endpoint: /api/v1/models/[model-id].json

Size: <1 KB per model (vs ~800 KB for full dataset)

# Get metadata for a specific model
curl https://llm-registry.dev/api/v1/models/claude-3-5-sonnet.json

# Get metadata for GPT-4o
curl https://llm-registry.dev/api/v1/models/openai-gpt-4o.json

✅ Benefit: Reduces API payload from ~800 KB to <1 KB per model request!

Rate Limiting (Cloudflare WAF)

Rate limiting is handled by Cloudflare WAF at the edge:

Limit

100 requests/min per IP

Enforcement

Cloudflare WAF (edge)

Response

HTTP 429 Too Many Requests

Note: Cloudflare returns HTTP 429 directly. No rate limit headers are sent by the application.

Cloudflare Dashboard Configuration

Go to Security → WAF → Rate limiting rules
Create rule: "API Rate Limiting"
Expression: (http.request.uri.path contains "/api/v1/")
Characteristics: ip.src
Request limit: 100 per minute
Mitigation: Block for 1 minute

Response Headers

Cache-Control

Responses are cached for 5 minutes (public, max-age=300)

Last-Modified

Date of the most recent score update

Attribution

All API responses include an attribution object. If you use this data, please credit:

Artificial Analysis — Scores marked with sourceId "artificial-analysis" are from artificialanalysis.ai
LLM Registry — Link back to llm-registry.com when displaying data