Alpha
Ctrl K
Contact
Support · NVIDIA Nemotron 3 Ultra

Will it last?

Ownership levelSubstantialnone·limited·partial·substantial·fullAnalytical input C ยท 63.6/100

This page is a projection of the one entry record, the Transparency factor that Support covers. The full verdict is set by all four factors together, floor-weighted so the weakest caps the whole.

Which domain expands which factor
  • AssessUse & modify + Transparency
  • ImplementData control + Reliability
  • UseReliability
  • SupportTransparency

Common problems & fixes

Aggregated pitfalls for a frontier-scale hybrid Mamba-Transformer MoE: multi-node MoE sharding and memory, kernel-to-engine version mismatches for the Mamba-2 and MoE layers, NVFP4 quantization-engine compatibility, and revision drift on download. Match engine and kernel versions to the documented stack, plan the sharding topology, pin the revision, and verify checksums. This is not an exhaustive catalogue.

Versions, changelog & cadence

Nemotron 3 Ultra was released around 2025-12 on the verified nvidia org with BF16 and NVFP4 variants. There is no single-checkpoint changelog; the Hugging Face revision hash is the anchor, so pin it.

Security / vulnerability disclosure

NVIDIA maintains a general corporate product-security reporting path (nvidia.com/en-us/security/). A model-specific vulnerability-disclosure policy for Nemotron was not confirmed this pass, so treat model-level issue reporting as going through the Hugging Face and GitHub org presence.

Community & support channels

Support is the NVIDIA open-model presence: the nvidia Hugging Face org (model discussions), the NVIDIA-NeMo GitHub organisation (the Nemotron recipes and NeMo Guardrails), and the NVIDIA/garak repository for red-team and safety-tooling questions.

Tracked known issues

From the card, recipe and white paper: the non-open intermediate checkpoints (not reproducible), the absence of a model-level safety evaluation, and the undisclosed training compute (systemic-risk status unresolved) are the recurring caveats. Track these against your own control stack and red-team rather than a published issue tracker.

How this scores

The ownership factor this domain covers, drawn from the one entry record.

2

TransparencyDo you know what it is: weights, training, behaviour, and legible terms?

Strong

open_weights_recipe: ungated weights, a white paper and a representative training recipe with CC-BY post-training data let you see a great deal of how Ultra was built. The limit is reproducibility, not visibility - the intermediate checkpoints and the 1M-context data are not open - which caps the openness tier, not transparency.

How this scores (AOI sub-dimensions)
Provenance3/5how well we can trace and verify what went into the modelDistributed from the verified nvidia org on Hugging Face as ungated safetensors with official BF16 / NVFP4 quantization variants and a clear canonical source (checklist ~4/8).
Governance3/5how accountable and well-documented the publisher isA reputable, accountable publisher (NVIDIA) with a white paper, a verified hub presence and an active release cadence, and a general corporate security-reporting path.
What this means for adoptionUltra is the one Nemotron 3 release whose ownership reaches substantial: the OpenMDW 'deal without restriction' grant makes use-and-modify strong, self-hosting via NIM makes data control strong, and the open_weights_recipe transparency is strong - so you can use it, see it and keep your data outright. It stops short of full because reliability is only moderate: there is no dedicated model-level safety evaluation, and at 550B / 55B active it is a multi-node deployment. Self-host, assemble the downloadable guard stack, red-team for your own use case, and treat EU systemic-risk status as unresolved until NVIDIA discloses training compute.

Sources

The same evidence records as the entry sheet. Read means the text was verified; unverified means it is known to exist but not yet read.

Model cardread2026-08-03
Nemotron 3 Ultra model card on the verified nvidia HF org: 550B total / 55B active; Mamba2-Transformer hybrid LatentMoE + MTP; up to 1M context; ~20T training tokens; ungated safetensors with BF16/NVFP4 variants; card states "open models with open weights, training data, and recipes"; licence tag openmdw-1.1.
Licenceread2026-08-03
OpenMDW License Agreement v1.1, read: "permission is hereby granted, free of charge, to deal in the Model Materials without restriction, including under all copyright, patent, database, and trade secret rights"; "Model Materials" covers weights, data, documentation and software.
Documentationread2026-08-03
Ultra training recipe at github.com/NVIDIA-NeMo/Nemotron (docs/nemotron/ultra3, Pretrain -> SFT -> MOPD -> Quant), read: described as "a representative single pass"; the "intermediate checkpoints it depends on have not been open-sourced" and the "1M-context LC phase is not included because its data ...
Technical_reportread2026-08-03
Nemotron 3 White Paper (arXiv 2512.20856), read: Ultra is a Mamba2-Transformer hybrid LatentMoE, up to 1M context.
Model cardread2026-08-03
Safety, read: no dedicated model-level safety evaluation for Ultra (white paper safety section is a contributor list; safety appears only as post-training data curation - Nemotron Content Safety v2 + Gretel refusal data + keyword/regex filtering).
Documentationread2026-08-03
Verified nvidia org on Hugging Face; the Ultra weight repo is ungated safetensors with published BF16/NVFP4 quantization variants.
Terms of serviceread2026-08-03
NVIDIA NIM is self-hosted deployment tooling ("data never leaves your secure enclave"), so NVIDIA is not an AOI inference-provider entry and self-hosting Ultra supports data control.
Third-party analysisread2026-08-03
NVIDIA is NOT a signatory to the EU GPAI Code of Practice (EU signatory list), and no NVIDIA training-content summary or copyright policy for Nemotron was located.