model_configuration.md
July 23, 2026 ยท View on GitHub
Back to Chat with your data README
Overview
Chat with Your Data uses Azure AI Foundry model deployments (a chat model, a reasoning model, and an embedding model) as its single inference surface. The backend calls Foundry with the workload's managed identity, so there are no model API keys to store. This guide lists the settings that control which models are deployed and how they are sized.
Available models
For the models and versions available in each region, see the Azure AI Foundry models documentation.
Default deployments
| Purpose | Model | Version | Deployment type | Capacity (TPM, thousands) |
|---|---|---|---|---|
| Chat | gpt-5.4-mini | 2026-03-17 | GlobalStandard | 50 |
| Reasoning | gpt-5-mini | 2025-08-07 | GlobalStandard | 50 |
| Embeddings | text-embedding-3-small | 1 | Standard | 100 |
Chat model parameters
| Parameter | Example | Purpose |
|---|---|---|
AZURE_GPT_MODEL_NAME | gpt-5.4-mini | Chat model name. |
AZURE_GPT_MODEL_VERSION | 2026-03-17 | Chat model version. |
AZURE_GPT_MODEL_DEPLOYMENT_TYPE | GlobalStandard | Chat model deployment type. |
AZURE_GPT_MODEL_CAPACITY | 50 | Tokens-per-minute limit (thousands). |
Reasoning model parameters
| Parameter | Example | Purpose |
|---|---|---|
AZURE_REASONING_MODEL_NAME | gpt-5-mini | Reasoning model name. |
AZURE_REASONING_MODEL_VERSION | 2025-08-07 | Reasoning model version. |
AZURE_REASONING_MODEL_DEPLOYMENT_TYPE | GlobalStandard | Reasoning model deployment type. |
AZURE_REASONING_MODEL_CAPACITY | 50 | Tokens-per-minute limit (thousands). |
Embedding model parameters
| Parameter | Example | Purpose |
|---|---|---|
AZURE_EMBEDDING_MODEL_NAME | text-embedding-3-small | Embedding model name. |
AZURE_EMBEDDING_MODEL_VERSION | 1 | Embedding model version. |
AZURE_EMBEDDING_MODEL_DEPLOYMENT_TYPE | Standard | Embedding model deployment type. |
AZURE_EMBEDDING_MODEL_CAPACITY | 100 | Tokens-per-minute limit (thousands). |
The embedding model sets the vector dimensions written to the retrieval index. text-embedding-3-small produces 1536-dimensional vectors. In PostgreSQL mode, the vector column width is set from this dimension when the index is first created; see PostgreSQL.
API version parameters
| Parameter | Example | Purpose |
|---|---|---|
AZURE_OPENAI_API_VERSION | 2025-01-01-preview | API version for chat and embedding calls. |
AZURE_AI_AGENT_API_VERSION | 2025-05-01 | API version for the Foundry agent runtime. |
Set and read parameters
Set a parameter before azd up:
azd env set <NAME> <VALUE>
Read the current values for your environment:
azd env get-values
For the full parameter list, including quota and region settings, see Customizing azd parameters and Model quota settings.