[andreatnvidia]

📋 Summary

NVIDIA deprecated the hosted Nemotron 3 Super endpoint on October 19, 2026, causing the built-in nvidia-reasoning health check to fail with ModelNotFoundError. Switch the alias to Super 3.5 VL with its supported high reasoning effort.

🔗 Related Issue

N/A - maintainer-requested fix for the failed provider health checks.

🔄 Changes

• Use nvidia/nemotron-3.5-super-vl-120b-a12b with reasoning_effort="high", retaining temperature 1.0 and top_p 0.95 per the replacement API reference.
• Update the existing default-model test to verify the model and inference parameters.
• Update the defaults reference and saved-configuration migration note. Existing configuration files remain preserved during upgrades.

🧪 Testing

• Config package suite: 830 passed.
• Ruff lint and formatting checks for the changed Python files.
• make check-fern-links DOCS_PYTHON=.venv/bin/python: 90 pages checked.
• Live provider health checks: 12 passed, 0 failed, including the new reasoning model. Run tested runtime commit 0386091e; the following commit changes documentation only.
• Independent Claude review: documentation finding addressed and follow-up confirmed no remaining actionable findings.

The full engine/interface test suites were not run for this model-default change.

✅ Checklist

• Follows commit message conventions.
• Commits are signed off (DCO).
• External-contributor issue requirement: N/A, maintainer-requested fix.
• Architecture documentation: N/A, model defaults only.

[greptile-apps]

Confidence Score: 5/5

[Medium impact] The PR appears safe to merge; no actionable issues were found.

Summary

Updates nvidia-reasoning to Nemotron Super 3.5 VL with reasoning_effort="high".

• The built-in NVIDIA reasoning alias now uses Nemotron 3.5 Super VL.

_{Reviews (2) · Last reviewed commit: "Merge branch 'main' into andreatnvidia/f..." · Reviewed by Greptile}

[github-actions]

Fern preview: https://nvidia-preview-pr-1127.docs.buildwithfern.com/nemo/datadesigner

Fern previews include the docs-website version archive with PR changes synced into latest. Notebook tutorials are rendered without execution outputs in previews.

[svcnvidia-nemo-ci]

No actionable findings. I reviewed all three changed files and traced configuration construction, provider request serialization, health checks, and saved-configuration initialization/reset paths.

The replacement model and reasoning_effort="high" agree across the built-in defaults, updated regression assertions, and defaults reference. The migration guidance matches the existing behavior of preserving saved configurations.

This was a static, read-only review. I didn't execute tests or independently verify the hosted endpoint's availability and accepted parameters.

Note: formal approval is not available for this request, so this review is posted as a comment only.

Review generated by codex. This was a full review. Request a full review with /review reset=true.

[nabinchha]

Thanks for putting this together, @andreatnvidia!

Summary

Updates nvidia-reasoning to Nemotron 3.5 Super VL with high reasoning effort, matching the stated intent.

Findings

No actionable findings in the three changed files at e8a621c2.

What Looks Good

• The regression test checks both the model and its complete inference parameters.
• Defaults and documentation agree with the NVIDIA API reference.
• The migration note explains saved-configuration preservation and reset behavior.

Residual Risk

The focused suite passed all 15 tests. Ruff lint and formatting passed for both changed Python files. Configuration construction, initialization, reset, and request serialization were inspected. Live endpoint calls and full package suites were not run.

Verdict

Ship it — no changes requested.

---

This review was generated by an AI assistant.