Use Custom Models

Overview

Future AGI supports two integration modes:

From Model Provider (Recommended): Direct integration with supported providers (OpenAI, AWS Bedrock, AWS SageMaker, Vertex AI, Azure), optimized for reliability, automatic updates, and simpler credential management.
Configure Custom Model (Advanced): Full flexibility to connect any model hosted behind an API endpoint, including in-house deployments, fine-tuned models, or proxy endpoints.

Once added, models are available platform-wide for custom evaluations.

Click here to learn how to create custom evaluations in Future AGI.

Future AGI currently supports:

Each provider has provider-specific authentication and cost configuration fields.
Set custom name to the model you are adding.
Provide input and output token costs for the model to compute cost when performing evaluations in Future AGI.

Use this when integrating self-hosted models, fine-tuned endpoints, or third-party APIs.

Field	About	Explanation	Example
Model Name	A friendly identifier for your model within Future AGI. This name appears in model selectors, dashboards, and evaluation reports.	Helps differentiate between multiple models, environments, and versions. Ensures better organization when running evaluations or RAG pipelines.	`mistral-rag-prod`
Input Token Cost per Million Tokens	The cost of input tokens (tokens sent in the request) per 1 million tokens.	Enables accurate billing visibility, cost attribution, and usage analytics within Future AGI dashboards.	`1.50` (represents $1.50 per 1M input tokens)
Output Token Cost per Million Tokens	The cost of output tokens (tokens generated in the response) per 1 million tokens.	Used to calculate total request costs alongside input tokens. Critical for cost optimization and reporting.	`2.00` (represents $2.00 per 1M output tokens)
API Base URL	The endpoint where Future AGI sends API requests to communicate with your custom model.	Required for model integration — Future AGI uses this endpoint for evaluations, RAG queries, prompt generation, and agent calls.	`https://api.my-model-server.com/v1`
Add Custom Configuration (Custom Key & Custom Value)	Lets you define custom headers, query parameters, or metadata required by your API.	Needed for scenarios like authentication, multi-tenant routing, model versioning, or passing provider-specific parameters.	Custom Key: `Authorization` Custom Value: `Bearer sk-123456`