Bring your own LLM

Configure custom OpenAI and Anthropic models for your AI agents in Cube Cloud.

Cube Cloud includes a set of predefined LLM models for powering your AI agents. If you need to use a specific model version, a fine-tuned model, or route requests through a particular inference provider, you can bring your own LLM (BYOLLM) by registering a custom model in Cube Cloud.

BYOLLM supports OpenAI and Anthropic model families. You can use models hosted directly by these providers or through compatible inference endpoints.

Supported providers

Cube Cloud supports the following LLM providers for custom models:

Provider	Description
OpenAI	Models accessed via the OpenAI API, including GPT-4o, GPT-4.1, and custom fine-tuned models
Anthropic	Models accessed via the Anthropic API, including Claude 3.5, Claude 3.7, Claude 4, and newer releases

Different inference providers that expose OpenAI-compatible or Anthropic-compatible APIs are also supported. This lets you route requests through providers such as Azure OpenAI Service, Amazon Bedrock, or any endpoint that implements the same API contract.

Register a custom model

To register a custom model, navigate to AI → Models in the Cube Cloud settings and add a new model entry. You will need to provide:

• Model name — a display name used to identify the model across your deployment
• Provider — select either OpenAI or Anthropic
• Model identifier — the model ID as recognized by the provider API (e.g., gpt-4o-2024-11-20, claude-sonnet-4-20250514)
• API key — the secret key for authenticating with the provider
• API base URL (optional) — a custom endpoint URL if you are using a third-party inference provider or a private deployment instead of the default provider endpoint

API keys are stored securely in Cube Cloud and are never exposed in logs or API responses. Rotate your keys regularly and use keys scoped to the minimum required permissions.

Assign a custom model to an agent

After registering a custom model, you can assign it to any agent.

Using the UI

In the Cube Cloud console, navigate to the agent configuration page and select your custom model from the LLM dropdown. Custom models appear alongside predefined models.

Using YAML configuration

Reference a custom model in your YAML agent configuration using the byom key:

agents:
  - name: my-agent
    space: my-space
    llm:
      byom:
        name: my-custom-model

You can also reference a custom model by its numeric ID:

agents:
  - name: my-agent
    space: my-space
    llm:
      byom:
        id: 123

Custom embedding models are also supported using the same syntax on the embedding_llm field:

agents:
  - name: my-agent
    space: my-space
    embedding_llm:
      byom:
        name: my-custom-embedding-model

Predefined models

Cube Cloud provides the following predefined models that you can use without any additional configuration:

Anthropic models:

Model ID	Description
claude_3_5_sonnetv2	Claude 3.5 Sonnet v2
claude_3_7_sonnet	Claude 3.7 Sonnet
claude_3_7_sonnet_thinking	Claude 3.7 Sonnet with extended thinking
claude_4_sonnet	Claude 4 Sonnet
claude_4_5_sonnet	Claude 4.5 Sonnet
claude_4_5_haiku	Claude 4.5 Haiku
claude_4_5_opus	Claude 4.5 Opus

OpenAI models:

Model ID	Description
gpt_4o	GPT-4o
gpt_4_1	GPT-4.1
gpt_4_1_mini	GPT-4.1 Mini
gpt_5	GPT-5
gpt_5_mini	GPT-5 Mini
o3	o3
o4_mini	o4-mini

Embedding models:

Model ID	Description
text-embedding-3-large	OpenAI text-embedding-3-large
text-embedding-3-small	OpenAI text-embedding-3-small

If none of the predefined models meet your requirements, use the BYOLLM feature to register a custom model.