Background texture
ModelMax

Providers

ModelMax is a unified gateway that routes your requests to upstream AI providers. You interact with a single API — ModelMax handles provider-specific authentication, formatting, and response normalization.

How routing works

When you send a request with a model parameter (e.g., deepseek-v3.2), ModelMax:

  1. Looks up the model in its registry.
  2. Identifies the upstream provider and provider-specific model name.
  3. Translates your OpenAI-compatible request into the provider's native format.
  4. Forwards the request and streams back the normalized response.

You don't need to know which provider serves which model — just use the model ID.

Available providers

Google Gemini

Google's Gemini API, accessed directly.

Models served: Gemini chat models and Veo video models.

Capabilities:

  • Full multimodal input (text, image, audio, video)
  • Streaming
  • Video generation (Veo)
  • Context caching

AWS Bedrock

Amazon's managed AI service, providing access to third-party models.

Models served: DeepSeek, Qwen, MiniMax, Kimi.

Capabilities:

  • Text chat with streaming
  • High availability through AWS infrastructure

Provider transparency

The owned_by field in the List Models response shows which provider serves each model:

{
  "id": "deepseek-v3.2",
  "owned_by": "bedrock"
}

No fallback

Each model is served by exactly one provider. ModelMax does not automatically fall back to a different provider if one is unavailable. If a provider has an issue, the API returns a 502 Bad Gateway error.

Adding providers

ModelMax is designed to add new providers without API changes. When new providers are added, their models appear in the model list and work with the same endpoints — no client-side changes needed.