Enterprise LLM Connection & Configuration Guide: AWS Bedrock, Azure AI Foundry, GCP Vertex AI & OpenAI Endpoints

Benchmarking foundation models across different cloud ecosystems requires uniform connection standards. Enterprise AI initiatives frequently compare models hosted on AWS Bedrock (Claude 3.7 Sonnet, Amazon Nova, Llama 3.3), Microsoft Azure AI Foundry (Azure OpenAI GPT-4o, DeepSeek R1, Phi-4), Google Cloud Vertex AI (Gemini 2.0 Flash/Pro), and Universal OpenAI / REST inference endpoints (vLLM, Ollama, self-hosted gateways). The LLM Evaluation Workbench requires verifying credentials via the Test Connection validation before initiating project benchmarking runs.

This guide explains exact URL endpoints, required authentication headers, model identifier formats, and step-by-step procedures for every supported provider.

1. AWS Bedrock Foundation Models

Amazon Bedrock provides unified access to high-performance foundation models with enterprise privacy and provisioned throughput.

Supported Model Identifiers

• anthropic.claude-3-7-sonnet-20250219-v1:0 — Premier reasoning and coding benchmark
• anthropic.claude-3-5-haiku-20241022-v1:0 — Ultra-low latency, cost-effective inference
• amazon.nova-pro-v1:0 — Multimodal reasoning and frontier enterprise intelligence
• amazon.nova-lite-v1:0 — High-throughput document and data extraction
• meta.llama3-3-70b-instruct-v1:0 — State-of-the-art open-weights instruction following
• amazon.titan-text-premier-v1:0 — Enterprise RAG and conversational foundation model

Configuration Requirements

• Target Provider: Select Bedrock (Amazon).
• AWS Region: Select your model deployment region (us-east-1, us-west-2, eu-central-1, or ap-southeast-1). Note that cross-region inference profiles (e.g. us.anthropic.claude-3-7-sonnet...) require regions with enabled Bedrock access.
• Authentication Header: Enter your Bedrock API token or AWS SigV4 authorization header:

  Bearer bedrock_ak_...

• System Prompt & Temperature: Optionally specify custom system instructions and temperature (default 0.2 for deterministic evaluations).

2. Microsoft Azure AI Foundry & Azure OpenAI

Azure AI Foundry allows provisioning models with custom deployment names, regional endpoints, and content safety filters.

Endpoint URL Format

Azure endpoints require the complete resource URL including model deployment and API version query parameters:

https://<your-resource-name>.services.ai.azure.com/models/chat/completions?api-version=2024-05-01-preview

Or Azure OpenAI Service endpoint:

https://<your-resource-name>.openai.azure.com/openai/deployments/<deployment-name>/chat/completions?api-version=2024-10-21

Configuration Requirements

• Target Provider: Select Microsoft Foundry (Azure AI).
• Model Identifier: Enter your deployment name or base model ID (e.g. gpt-4o, DeepSeek-R1, Phi-4).
• API Key / Auth Header: Azure accepts direct subscription keys:

  api-key: your-azure-foundry-api-key-here

  Or enter the 32-character key directly into the API Key input field.

3. Google Cloud Vertex AI (Gemini 2.0 & Claude on Vertex)

Vertex AI hosts Google's Gemini models and partner models with deep enterprise VPC and IAM integration.

Endpoint URL Format

https://<region>-aiplatform.googleapis.com/v1/projects/<project-id>/locations/<region>/publishers/google/models/<model-id>:generateContent

Examples of <model-id>: gemini-2.0-flash, gemini-2.0-pro-exp, gemini-1.5-pro.

Configuration Requirements

• Target Provider: Select Vertex AI (Google).
• Authentication Header: Google Cloud OAuth2 Access Token:

  Bearer ya29.a0AfH...

  Generate an ephemeral token in your terminal via:

  gcloud auth print-access-token

4. Universal OpenAI & REST Inference Gateways

Evaluate any service adhering to the OpenAI-compatible /v1/chat/completions specification, including OpenAI, Anthropic, Mistral, Groq, Ollama, and vLLM clusters.

Configuration Requirements

• Target Provider: Select OpenAI / REST (Universal).
• Endpoint URL:
  • OpenAI Official: https://api.openai.com/v1/chat/completions
  • Ollama Local Inference: http://localhost:11434/v1/chat/completions
  • vLLM Server: http://your-gpu-server:8000/v1/chat/completions
  • Groq: https://api.groq.com/openai/v1/chat/completions
• Model Identifier: e.g. gpt-4o, gpt-4o-mini, llama-3.3-70b-versatile, mistral-large-latest.
• API Key: Standard API key (e.g. sk-proj-... or Bearer <token>).

Connection Validation Diagnostic Matrix

Ready to benchmark foundation models? Visit the LLM Evaluation Workbench, click New Project, run Test Connection to validate your settings, and launch your evaluation suite.