Supported LLM providers and models
Which LLM providers and models the AI services of the Arango Contextual Data Platform support, how OpenAI-compatible endpoints are reached, and where each service is configured
The AI services of the Arango Contextual Data Platform work with
OpenAI-compatible APIs. The recommended setup is the openai provider with the
OpenAI models listed below: that is the combination ArangoDB tests, and other
endpoints can differ in behavior such as latency.
You can still use any other OpenAI-compatible endpoint (OpenRouter, Google
Gemini, Anthropic, Azure, or a corporate LLM) and run a model that is not on the
list. In the Importer, AutoGraph, and AutoRAG, use the custom provider for
these: it is the intended way to point a service at an OpenAI-compatible
endpoint that is not the OpenAI API itself, and you should always set
chat_api_url / embedding_api_url explicitly with it. In those three
services, pointing the openai provider at a non-OpenAI URL is not
supported. Natural Language to AQL has no custom provider and reaches such
endpoints with openai plus a chat_api_url, see
Natural Language to AQL setup. Models
beyond the ones listed below are outside ArangoDB’s testing, so validate them in
your own environment.
“OpenAI-compatible” here has a specific meaning: the services talk to providers
through the OpenAI Chat Completions client, so an endpoint must implement the
/v1/chat/completions contract that client expects (and /v1/embeddings for
embedding models). An endpoint that exposes only a different API surface is not
supported, even if it is marketed as OpenAI-compatible. Some newer OpenAI models
require the Responses API (/v1/responses) instead; the Importer and AutoGraph
detect this and fall back automatically.
A model is listed as supported only if it works seamlessly across the Importer, AutoRAG, and AutoGraph services. Individual services may also work with additional models. For the full list available to a specific service, see that service’s own documentation (for example, Importer LLM Configuration).
| Model | Type | Services | Default |
|---|---|---|---|
gpt-5.4-mini | Chat (LLM) | Importer, AutoGraph, AutoRAG, Ada | — |
gpt-5.4-nano | Chat (LLM) | Importer, AutoGraph, AutoRAG, GraphRAG | Yes |
gpt-5 | Chat (LLM) | Importer, AutoGraph, AutoRAG, Natural Language to AQL, Ada | — |
gpt-5-mini | Chat (LLM) | Importer, AutoGraph, AutoRAG, Ada | — |
gpt-5-nano | Chat (LLM) | Importer, AutoGraph, AutoRAG, Ada | — |
o3 | Chat (LLM) | Importer, AutoGraph, AutoRAG, Ada | — |
text-embedding-3-small | Embedding | Importer, AutoGraph, AutoRAG, GraphRAG | Yes |
text-embedding-3-large | Embedding | Importer, AutoGraph, AutoRAG | — |
A model marked Yes under Default is the one applied automatically when the
model name is not set with the openai provider. Whether the custom provider
falls back to the same model varies by service — check the parameter reference
for the service you are configuring.
Where each service is configured
The provider and model are set per service, because their parameters differ:
API keys are stored once in the Secrets Manager and selected in each service’s configuration.