Models

List available models and their metadata using GET /v1/models.

Endpoint

GET /v1/models

Returns the list of models currently registered in the gateway in OpenAI-compatible format. Each model entry includes gateway-specific metadata such as provider, model type, reasoning tier, streaming support, and embedding dimensions where relevant.

Response

{
  "object": "list",
  "data": [
    {
      "id": "llama-3.3-70b-versatile",
      "object": "model",
      "created": 1714000000,
      "owned_by": "groq",
      "type": "chat",
      "provider": "groq",
      "reasoning_tier": null,
      "supports_streaming": true
    },
    {
      "id": "gemini-2.0-flash",
      "object": "model",
      "created": 1714000000,
      "owned_by": "gemini",
      "provider": "gemini",
      "reasoning_tier": null,
      "supports_streaming": true
    },
    {
      "id": "gemini-2.0-flash-thinking-exp",
      "object": "model",
      "created": 1714000000,
      "owned_by": "gemini",
      "provider": "gemini",
      "reasoning_tier": "high",
      "supports_streaming": true
    },
    {
      "id": "@cf/meta/llama-3.3-70b-instruct-fp8-fast",
      "object": "model",
      "created": 1714000000,
      "owned_by": "workers_ai",
      "provider": "workers_ai",
      "reasoning_tier": null,
      "supports_streaming": true
    },
    {
      "id": "voyage-3",
      "object": "model",
      "created": 1714000000,
      "owned_by": "voyage_ai",
      "type": "embedding",
      "provider": "voyage_ai",
      "reasoning_tier": null,
      "supports_streaming": false,
      "dimensions": 1024,
      "enabled": true
    }
  ]
}

Response Fields

Continuous evals are optional and never bypass hard filters. The router still excludes unhealthy, rate-limited, incompatible, or capability-mismatched models before applying eval weights. Embedding rows are included so RAG services can choose vector models from the live gateway catalog. Embedding vector spaces are not interchangeable; persist the selected model with each index.

Model Capabilities

Each model has capability metadata used for automatic routing:

When your request requires a capability, the gateway automatically filters to models that support it. If no capable model is available, a 503 is returned.

Capability Summary by Provider

Examples

curl https://your-gateway.workers.dev/v1/models \
  -H "Authorization: Bearer <GATEWAY_API_KEY>"

const response = await fetch('https://your-gateway.workers.dev/v1/models', {
  headers: {
    'Authorization': 'Bearer <GATEWAY_API_KEY>',
  },
});

const { data: models } = await response.json();

// Print all model IDs grouped by provider
const byProvider = models.reduce((acc, model) => {
  (acc[model.provider] ??= []).push(model.id);
  return acc;
}, {});

for (const [provider, ids] of Object.entries(byProvider)) {
  console.log(`\n${provider}:`);
  ids.forEach((id) => console.log(`  - ${id}`));
}

import OpenAI from 'openai';

const client = new OpenAI({
  baseURL: 'https://your-gateway.workers.dev/v1',
  apiKey: '<GATEWAY_API_KEY>',
});

const models = await client.models.list();

// Filter to only streaming-capable chat models
const chatModels = models.data.filter(
  (m) => m.type === 'chat' && m.supports_streaming,
);

console.log('Available chat models:');
chatModels.forEach((m) => console.log(`  ${m.id} (${m.provider})`));

Filtering by Provider

You can filter client-side after fetching the full list:

const response = await fetch('https://your-gateway.workers.dev/v1/models', {
  headers: { 'Authorization': 'Bearer <GATEWAY_API_KEY>' },
});

const { data: models } = await response.json();

const groqModels = models.filter((m) => m.provider === 'groq');
const reasoningModels = models.filter((m) => m.reasoning_tier !== null);
const embeddingModels = models.filter((m) => !m.supports_streaming);

Using a Specific Model

Pass any id value directly as the model field in a chat or embeddings request:

curl https://your-gateway.workers.dev/v1/chat/completions \
  -H "Authorization: Bearer <GATEWAY_API_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "llama-3.3-70b-versatile",
    "messages": [{ "role": "user", "content": "Hello from Groq!" }]
  }'

Or use "auto" to let the gateway pick the healthiest available provider automatically.