[ZedingZhang (ZedingZhang)]

Motivation & Context

FoundryEvals() passes "gpt-4o" as an explicit model when creating its own FoundryChatClient. This silently overrides FOUNDRY_MODEL, so evaluations use the default deployment even when another deployment is configured.

Description & Review Guide

• What are the major changes? Resolve the model through the existing load_settings and FoundrySettings before applying the "gpt-4o" fallback. Document the precedence and add seven regression cases covering explicit values, environment values, unset/empty environment values, and both injected-client paths.
• What is the impact of these changes? Auto-created clients now use an explicit model, then FOUNDRY_MODEL, then "gpt-4o". Model resolution for injected clients keeps its existing behavior.
• What do you want reviewers to focus on? The model fallback is confined to the auto-created-client branch. The environment-only regression fails on the original implementation and passes with the fix.

Related Issue

Fixes #9951. The issue has the ready-for-implementation label; no competing open PR was found before submission.

AI Assistance

• No material AI assistance was used.
• This is an AI-assisted contribution.

AI assistance details: Codex assisted with issue investigation, implementation, regression tests, the docstring update, and this PR description. Codex executed the reported checks; I reviewed all submitted content and take responsibility for it.

Contribution Checklist

• The code builds clean without any errors or warnings — Foundry sdist and wheel built successfully.
• All unit tests pass, and I have added new tests where possible — Foundry unit suite: 473 passed, 29 integration tests deselected; seven new regression cases. Ruff formatting/lint, strict source Pyright, and all five test type checkers also pass. No live Foundry integration tests were run.
• The PR follows the Contribution Guidelines
• This PR links to an agreed issue with no competing open PR, or the Related Issue section documents a trivial-change or repository-automation exception.
• This is not a breaking change. If it is a breaking change, add the breaking change label (or add "[BREAKING]" to the title prefix, before or after any language prefix) — a workflow keeps the label and title prefix in sync automatically.

[Copilot]

🟢 Approval recommended

No active review issues are recorded for the change that makes auto-created FoundryEvals clients honor FOUNDRY_MODEL.

No open issues 🎉

What changed in this PR

Fixes #9951 by making auto-created FoundryEvals clients respect FOUNDRY_MODEL instead of overriding it with the default deployment.

Changes:

• Resolves the model in order: explicit argument, FOUNDRY_MODEL, then "gpt-4o".
• Documents the fallback behavior while preserving injected-client behavior.
• Adds seven regression cases covering model precedence, unset/empty environment values, and injected clients.

File	Description
python/​packages/​foundry/​tests/​test_foundry_evals.py	Adds model-precedence and injected-client regression tests.
python/​packages/​foundry/​agent_framework_foundry/​_foundry_evals.py	Resolves settings before applying the default model and documents the behavior.

_{Review effort: Balanced}