Skip to main content
Returns a list of vision-capable models available through OpenRouter. Results are cached for 1 hour to minimize API calls.

​
Endpoint

GET /api/models

​
Use Cases

  • Discover available vision models for OCR
  • Compare model pricing and capabilities
  • Dynamically populate model selection dropdowns
  • Check context lengths and pricing before processing

​
Request

No parameters required. Simple GET request:
cURL
curl http://localhost:3000/api/models
JavaScript
const response = await fetch('/api/models');
const data = await response.json();
console.log(`Found ${data.count} vision models`);
Python
import requests

response = requests.get('http://localhost:3000/api/models')
data = response.json()
print(f"Found {data['count']} vision models")

​
Response

​
models
VisionModel[]
required
Array of vision-capable models
​
cached
boolean
required
Whether this response was served from cache
​
count
number
Total number of models returned (only on fresh fetch)
​
cacheAge
number
Age of cached data in seconds (only when cached=true)

​
Example Response

{
  "models": [
    {
      "id": "google/gemini-2.0-flash",
      "name": "Google: Gemini 2.0 Flash",
      "description": "Fast multimodal model with vision capabilities",
      "contextLength": 1000000,
      "pricing": {
        "promptPer1k": 0.075,
        "completionPer1k": 0.30,
        "imagePer1k": 0.01315789
      },
      "provider": "google",
      "isVision": true
    },
    {
      "id": "openai/gpt-4o",
      "name": "OpenAI: GPT-4o",
      "description": "High-intelligence flagship model for complex tasks",
      "contextLength": 128000,
      "pricing": {
        "promptPer1k": 2.50,
        "completionPer1k": 10.00,
        "imagePer1k": 7.225
      },
      "provider": "openai",
      "isVision": true
    },
    {
      "id": "anthropic/claude-3.5-sonnet",
      "name": "Anthropic: Claude 3.5 Sonnet",
      "description": "Highest level of intelligence and capability",
      "contextLength": 200000,
      "pricing": {
        "promptPer1k": 3.00,
        "completionPer1k": 15.00,
        "imagePer1k": 4.80
      },
      "provider": "anthropic",
      "isVision": true
    }
  ],
  "cached": false,
  "count": 3
}

​
Caching Behavior

The endpoint caches results for 1 hour (3600 seconds) to reduce load on OpenRouter’s API. Fresh Response: 
{
  "models": [...],
  "cached": false,
  "count": 45
}
Cached Response: 
{
  "models": [...],
  "cached": true,
  "cacheAge": 1823
}
Cache is server-side and shared across all requests. First request after server restart or cache expiry will fetch fresh data.

​
Vision Model Detection

The endpoint filters models based on multiple criteria:
  1. Architecture modality: Checks for “image”, “vision”, or “multimodal” in model architecture
  2. Image pricing: Models with non-zero image pricing are considered vision-capable
  3. Model name patterns: Detects common vision model identifiers:
    • GPT-4o, GPT-4-turbo
    • Gemini (all versions)
    • Claude 3+ (Opus, Sonnet, Haiku)
    • LLaVA, Pixtral, Qwen-VL
    • CogVLM, InternVL, Yi-Vision

​
Sorting

Models are sorted by:
  1. Provider name (alphabetically)
  2. Model name (alphabetically within provider)
Example order:
  • anthropic/claude-3-haiku
  • anthropic/claude-3.5-sonnet
  • google/gemini-2.0-flash
  • openai/gpt-4o-mini

​
Use with OCR Endpoints

Select a model from this list to use with OCR endpoints: 
// 1. Fetch available models
const modelsResponse = await fetch('/api/models');
const { models } = await modelsResponse.json();

// 2. Find a suitable model
const gemini = models.find(m => m.id === 'google/gemini-2.0-flash');

// 3. Use it for OCR
const ocrResponse = await fetch('/api/ocr-structured-v4', {
  method: 'POST',
  headers: { 'Content-Type': 'application/json' },
  body: JSON.stringify({
    imageBase64: '...',
    model: gemini.id
  })
});

​
Error Responses

{
  "error": "Failed to fetch models: 401 Unauthorized",
  "models": []
}
The server failed to fetch models from OpenRouter. Check server logs for details.
{
  "error": "Network request failed",
  "models": []
}
Network connectivity issue. Returns cached models if available, otherwise empty array.

​
Implementation Details

Source: app/api/models/route.ts Cache Implementation:
  • In-memory cache (resets on server restart)
  • 1-hour TTL (3600000ms)
  • Atomic cache updates
Filtering Logic:
  • Fetches all models from OpenRouter
  • Filters to vision-capable only
  • Transforms to simplified schema
  • Sorts by provider then name

Model Selection Guide

Learn how to choose the right model

OCR Endpoints

Use models with OCR extraction

Request Formats

Override model per request

Compare Page

Test multiple models side-by-side

Build docs developers (and LLMs) love

Get started for freeTalk to us