List LLM Models

Use the /v3/models endpoint to retrieve all LLM models available through Eden AI, along with their capabilities: input modalities, reasoning, web search, and function calling.

Endpoint

GET /v3/models

The endpoint is public: no API key is required.

Example

curl https://api.edenai.run/v3/models

import requests

response = requests.get("https://api.edenai.run/v3/models")

for model in response.json()["data"]:
    print(model["id"])

const response = await fetch("https://api.edenai.run/v3/models");

const { data } = await response.json();
data.forEach(model => console.log(model.id));

Response

Each entry includes the model’s identity, context window, capabilities, pricing, available regions, and whether the provider retains your data (list truncated for readability):

{
  "object": "list",
  "data": [
    {
      "id": "google/gemini-flash-latest",
      "object": "model",
      "created": 1784646733,
      "owned_by": "google",
      "model_name": "gemini-3.6-flash",
      "context_length": 1048576,
      "description": "Gemini 3.6 Flash is a high-efficiency model from Google...",
      "capabilities": {
        "input_modalities": ["text", "image", "video", "file", "audio"],
        "output_modalities": ["text"],
        "supports_reasoning": true,
        "supports_web_search": true,
        "supports_tool_choice": true,
        "supports_computer_use": false,
        "supports_prompt_caching": true,
        "supports_response_schema": true,
        "supports_system_messages": true,
        "supports_function_calling": true,
        "supports_native_streaming": true,
        "supports_assistant_prefill": false,
        "supports_embedding_image_input": false,
        "supports_parallel_function_calling": true
      },
      "pricing": {
        "input_cost_per_token": 1.5e-06,
        "output_cost_per_token": 7.5e-06,
        "cache_read_input_token_cost": 1.5e-07,
        "cache_creation_input_token_cost": 8.33e-08
      },
      "regions": [
        { "code": "eu", "name": "Europe" }
      ],
      "zero_data_retention": false,
      "alias_of": "google/gemini-3.6-flash"
    }
  ]
}

Each model id is used directly as the model parameter in your requests. pricing holds the rates a call bills at, after any discount, and list_pricing the provider’s list rates before it, most of them in US dollars per token. OpenAI models that can run as a batch also carry input_cost_per_token_batches, output_cost_per_token_batches and cache_read_input_token_cost_batches, the rates a batch bills at when the model has them.

Grouped view: one entry per routable name

Many models are sold by more than one provider. Pass view=models to get one entry per routable model name, with the provider endpoints nested underneath. This is how you discover the names that Provider Routing accepts.

cURL

curl "https://api.edenai.run/v3/models?view=models" \
  -H "Authorization: Bearer YOUR_API_KEY"

The key is optional here: without one, the listing holds the catalog only.

{
  "object": "list",
  "data": [
    {
      "id": "gpt-5.6-sol",
      "object": "model",
      "created": 1785352476,
      "owned_by": "openai",
      "mode": "chat",
      "endpoint_count": 2,
      "context_length": 131072,
      "capabilities": {
        "...": "what at least one endpoint supports"
      },
      "pricing": {
        "...": "full price list of pricing_endpoint"
      },
      "pricing_endpoint": "openai/gpt-5.6-sol",
      "regions": [
        { "code": "eu", "name": "Europe" }
      ],
      "endpoints": [
        { "id": "openai/gpt-5.6-sol", "owned_by": "openai", "...": "a full model entry, as above" },
        { "id": "azure/gpt-5.6-sol",  "owned_by": "azure",  "...": "a full model entry, as above" }
      ]
    }
  ]
}

What to expect from the shape:

id is the routable name. Send it as model and Eden AI picks which provider serves it.
Top-level fields summarise the model across its endpoints. context_length is the largest window among the endpoints. capabilities is the union across endpoints, so a flag is true and a modality listed when any endpoint supports it. pricing carries the full price list of the cheapest endpoint for a typical call (untagged endpoints only, since an @region variant prices a tagged call), identified by pricing_endpoint; both are null when no endpoint has a price. A routed call is billed at the prices of the provider that served it. regions lists every region where at least one endpoint runs, sorted by code.
Each entry in endpoints is the same object the default view returns, with that provider’s own pricing, capabilities, context length, and regions. To compare providers or see per-provider values, read endpoints.
owned_by changes meaning between the views. On a grouped entry it is who authored the model (openai); on an endpoint entry it is the provider selling it (azure).
created is the newest endpoint’s timestamp, and endpoint_count equals endpoints.length.
The default (view=endpoints, or no parameter) is unchanged, and both entry shapes stay compatible with the OpenAI SDK’s client.models.list().
Your routing policies are listed here too. Send your API key and the grouped view adds one entry per routing policy you can call, under its policy/{name} id, with owned_by set to edenai, an empty endpoints list and endpoint_count at 0: which target serves a call is decided per request, so pricing, context_length and regions are null or empty. Policies never appear in the default view or under a zero_data_retention filter.

Every model listing takes the same parameter: /v3/models, /v3/embeddings/models, /v3/moderations/models, /v3/audio/transcriptions/models, /v3/audio/speech/models, /v3/images/models, /v3/videos/models and /v3/rerank/models.

Zero Data Retention

Every endpoint entry (each entry of the default view, and each item of endpoints with view=models) carries zero_data_retention, which says whether that provider keeps your prompts and outputs after the response, for the calls Eden AI sends it:

Value	Meaning
`true`	The provider keeps no prompts or outputs after the response (Zero Data Retention)
`false`	The provider keeps them, for example for abuse monitoring
`null`	Not verified yet

The value is set per provider, and per model when one model’s terms differ from its provider’s. Models served from your own private deployment are always true. See Provider Data Policies for how each provider handles your data. To list only Zero Data Retention endpoints, pass zero_data_retention=true:

cURL

curl "https://api.edenai.run/v3/models?zero_data_retention=true"

zero_data_retention=false lists only the endpoints known to keep data, and entries still at null match neither. With view=models, each model keeps only its matching endpoints and recomputes its model-level fields from them, and is left out when none match. Every model listing above takes the parameter.

Model aliases

Some models are available under a stable alias, a version-agnostic name that always points to the current release, alongside versioned or dated snapshot IDs:

Stable alias: google/gemini-flash-latest, anthropic/claude-sonnet-latest, deepseek/deepseek-chat
Versioned snapshot: anthropic/claude-opus-4-5-20251101

Use the stable alias when you want your integration to keep working as providers ship new versions. Pin a versioned snapshot when you need a fixed, reproducible model. Either form works as the model parameter. Alias entries report the concrete model they currently resolve to in the alias_of field, as in the response example above. The catalog lists both forms when a provider exposes them, so you may see a stable alias (e.g. anthropic/claude-opus-4-5) and one or more dated snapshots (e.g. anthropic/claude-opus-4-5-20251101) side by side. Use the stable alias to always get the latest version, or a dated snapshot to pin to a specific release.

Capabilities

The capabilities object describes what each model supports:

Field	Description
`input_modalities`	Input types the model accepts: `text`, `image`, `audio`, `video`, `file`
`output_modalities`	Output types the model produces
`supports_reasoning`	Model supports extended thinking / reasoning mode
`supports_web_search`	Model can perform live web searches via `web_search_options`
`supports_function_calling`	Model supports function/tool calling
`supports_tool_choice`	Model supports the `tool_choice` parameter
`supports_prompt_caching`	Model supports prompt caching
`supports_response_schema`	Model supports structured output with a response schema
`supports_system_messages`	Model accepts system messages
`supports_computer_use`	Model supports computer-use tooling
`supports_parallel_function_calling`	Model can call multiple tools in one turn
`supports_assistant_prefill`	Model accepts a pre-filled assistant message
`supports_native_streaming`	Model supports native streaming responses
`supports_embedding_image_input`	Model accepts image input for embeddings

Some models expose extra provider-specific keys (e.g. supports_pdf_input, supports_audio_input), and a few return "capabilities": null. Treat a missing key or a null object as “not supported”.

You can also browse all features and providers visually in the Eden AI model catalog.

Need EU data residency? Hit https://api.eu.edenai.run/v3/models instead and the list is automatically filtered to EU-eligible models. See EU Endpoint.

Looking for OCR, image, or audio models? See List Expert Models for the full catalog of expert model features.

V3 Documentation

Quick Start

Overview

LLMs

Expert Models

General

Organization

Data Governance

Integrations

Endpoint

Example

Response

Grouped view: one entry per routable name

Zero Data Retention

Model aliases

Capabilities

​Endpoint

​Example

​Response

​Grouped view: one entry per routable name

​Zero Data Retention

​Model aliases

​Capabilities

Endpoint

Example

Response

Grouped view: one entry per routable name

Zero Data Retention

Model aliases

Capabilities