Forum Discussion

Hariteja2189's avatar
Hariteja2189
Copper Contributor
Nov 04, 2025

Less models in ai foundry that supports agentic use

Hi,
I have seen that nearly 11,000 models are available in Azure ai foundry, but when I try to deploy models that support Agents, only 18 models are available for selection. Is there any reason behind this ? Are we planning to support many models from external providers or rely on gpt models as first priority

ai agents
ai foundry

1 Reply

  • Surya_Narayana's avatar
    Surya_Narayana
    MCT
    Nov 19, 2025

    hi Hariteja2189​ Great question — this is a common point of confusion when moving from the general model catalog to the Agents feature within Azure AI Foundry.

    Here’s the real reason:

    Only Certain Models Are Designed for “Agentic” Workloads

    Agentic features (tools, function-calling, planning, multi-step reasoning, memory, tool orchestration, etc.) require models that support:

    • Function calling / tool calling
    • Structured outputs
    • Deterministic reasoning APIs
    • Long context support
    • Agent control signals
    • Safe + predictable action execution
    • Azure governance requirements

    Most of the 11,000+ models in the model catalog (HuggingFace, Open source, Community, Research) do not implement these capabilities.

    That’s why only a subset of enterprise-ready models can be used to build Agents.

    Why So Few Models Right Now?

    Because Agentic systems require:

    Strict security

    Predictable function behavior

    Tool compatibility

    Azure policy compliance

    Safety alignment

    Enterprise-grade reliability

    Most community or open models do not meet these requirements yet.

    Microsoft has intentionally started with ~18 models that are:

    • fully validated
    • safe for enterprise agents
    • tested to support tool-calling
    • optimized for long-running, multi-step agent tasks

     

    Are More Models Coming? — YES

    Yes — Microsoft is already working on expanding support.

    Expect support to grow for:

    🔹 Llama-3.1 agentic-capable models

    🔹 Phi-3 / Phi-4 updates

    🔹 Mistral tool-calling models

    🔹 Meta’s future agentic-tuned models

    🔹 Additional Azure OpenAI models

    🔹 Select vendor-hosted third-party models (when they meet agentic requirements)

    But every model must first pass:

    • safety checks
    • function-calling interoperability
    • agent policy compliance
    • latency and reliability validation

    so rollout is incremental, not bulk.

     

     

     

Resources