Step 4: LiteLLM Proxy Model Configuration
The core of the LiteLLM Proxy configuration is the model list, which defines the LLM models the proxy will manage. Sample configuration files are provided in the Helm chart for each major cloud provider.
Configuration File Locations
- AWS Bedrock:
litellm/config/litellm-aws-config.yaml - Azure OpenAI:
litellm/config/litellm-azure-config.yaml - Google Vertex AI:
litellm/config/litellm-gcp-config.yaml
Regional Configuration Overview
The sample configurations specify default regions for accessing the models. The table below provides a summary.
| Provider | Model | Default Region |
|---|---|---|
| AWS Bedrock | claude-3-5-sonnet | us-west-2 |
| AWS Bedrock | claude-3-5-sonnet-v2 | us-west-2 |
| AWS Bedrock | claude-3-7 | us-west-2 |
| AWS Bedrock | claude-4-sonnet-1m | us-west-2 |
| AWS Bedrock | claude-4-sonnet | us-west-2 |
| AWS Bedrock | claude-4-sonnet | eu-central-1 |
| AWS Bedrock | claude-4-opus | us-west-2 |
| AWS Bedrock | claude-4-1-opus | us-west-2 |
| AWS Bedrock | titan | us-west-2 |
| Google Vertex AI | gemini-2.0-flash | us-central1 |
| Google Vertex AI | gemini-2.5-pro | us-central1 |
| Google Vertex AI | claude-sonnet-v2-vertex | us-east5 |
| Google Vertex AI | claude-sonnet-3-7-vertex | us-east5 |
| Google Vertex AI | text-embedding-005 | us-central1 |
| Azure OpenAI | gpt-4o-2024-08-06 | eastus2 |
| Azure OpenAI | gpt-4o-2024-11-20 | eastus2 |
| Azure OpenAI | gpt-4o-mini | eastus2 |
| Azure OpenAI | gpt-4.1 | eastus2 |
| Azure OpenAI | gpt-4.1-mini | eastus2 |
| Azure OpenAI | gpt-5-2025-08-07 | eastus2 |
| Azure OpenAI | gpt-5-mini-2025-08-07 | eastus2 |
| Azure OpenAI | gpt-5-nano-2025-08-07 | eastus2 |
| Azure OpenAI | gpt-4-vision | eastus2 |
| Azure OpenAI | o3-mini | eastus2 |
| Azure OpenAI | o1 | eastus2 |
| Azure OpenAI | o3-2025-04-16 | eastus2 |
| Azure OpenAI | o4-mini-2025-04-16 | eastus2 |
| Azure OpenAI | codemie-text-embedding-ada-002 | eastus2 |
Customizing Regions
You can customize the regions where your models are accessed by modifying the corresponding configuration files.
AWS Bedrock
To use a different AWS region, modify the aws_region_name parameter in the model's configuration.
model_list:
- model_name: claude-4-sonnet
litellm_params:
aws_region_name: us-west-2 # Change this to your preferred region, e.g., "eu-central-1"
Azure OpenAI
For Azure, the region is determined by the endpoint URL in the api_base parameter.
model_list:
- model_name: gpt-4o-2024-08-06
litellm_params:
model: azure/gpt-4o-2024-08-06
api_base: https://your-resource.openai.azure.com/ # Your Azure OpenAI endpoint defines the region
litellm_credential_name: default_azure_openai_credential
Google Vertex AI
For Google Vertex AI, the method for region configuration depends on the model family:
Gemini Models
Region is set globally in the litellm_settings section.
litellm_settings:
vertex_project: os.environ/VERTEX_PROJECT
vertex_location: 'us-central1' # Change this to your preferred region for Gemini models
Claude Models on Vertex AI
Region and project are specified directly in each model's configuration.
model_list:
- model_name: claude-sonnet-3-7-vertex
litellm_params:
model: vertex_ai/claude-3-7-sonnet
vertex_ai_project: os.environ/VERTEX_PROJECT
vertex_ai_location: 'us-east5' # Change this to your preferred region for this specific model
Next Steps
Continue to Deployment.