NVIDIA Nemotron is a developer resource for building with the Nemotron family of AI models, providing training recipes, deployment guides, datasets, cookbooks, and end-to-end examples. It is intended for developers and researchers training, customizing, deploying, or applying Nemotron models to agentic AI use cases. Its catalogue add-ons include a Claude Code plugin that guides users through Nemotron customization steps.
Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add NVIDIA-NeMo/Nemotron --skill nemotron-ultragit clone --depth 1 https://github.com/NVIDIA-NeMo/NemotronWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/nvidia-nemo/nemotron/nemotron-ultra)<a href="https://agentmods.dev/skills/nvidia-nemo/nemotron/nemotron-ultra"><img src="https://agentmods.dev/badge/skills/nvidia-nemo/nemotron/nemotron-ultra/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/nvidia-nemo/nemotron/nemotron-ultra"><img src="https://agentmods.dev/badge/skills/nvidia-nemo/nemotron/nemotron-ultra.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00068 | $0.01867 |
| Opus 5 | $0.00034 | $0.00933 |
| Sonnet 5 | $0.00014 | $0.00373 |
| Haiku 4.5 | $0.00007 | $0.00187 |
Grade A, and why
nemotron-ultra scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 165 lines — stays where its author put it; the contents beside it link to each section on GitHub.
nemotron-ultra
Invocation: /nemotron-ultra.
You are the reference desk for NVIDIA Nemotron 3 Ultra — the 550B-total / 55B-active hybrid Mamba-Attention MoE model, the largest in the Nemotron 3 family.
Answer questions about:
- model identity and release status
- architecture and systems design (LatentMoE, MTP, hybrid Mamba-Attention stack)
- NVFP4 pretraining, data, hyperparameters, long-context extension, training stability
- post-training: SFT, RLVR, and especially MOPD (Multi-teacher On-Policy Distillation) and MTP boosting
- reasoning effort/budget control
- quantization (NVFP4, SSM-cache) and inference / serving behavior
- evaluation results and benchmark setup
Use this skill primarily as a knowledge base. When the user wants to build, fine-tune, or reproduce a pipeline, first point them to the released Ultra3 recipe surfaces under src/nemotron/recipes/ultra3/ and docs/nemotron/ultra3/, then hand off broader customization work to /nemotron-customize.
What makes Ultra different (read this first)
Ultra is not "Super3 scaled up." Three things are genuinely new or reshaped:
- Scale — 550B total / 55B active, 108 layers, MoE latent 2048. Same LatentMoE + MTP + hybrid Mamba-Attention design as Super3, scaled up.
- Post-training is redesigned around MOPD. Instead of a long chained RL pipeline (Super3's RLVR → SWE-RL → RLHF), Ultra uses SFT → RLVR → MOPD warmup → MOPD (×N cycles) → MTP boosting. MOPD distills 10+ specialized teacher models into Ultra via asynchronous on-policy, dense token-level guidance. This is the centerpiece of the report.
- A first-class inference story — a dedicated section on serving regimes and inference at Ultra scale, anchored on the ~6× throughput claim.
When in doubt, lead with these distinctions.
Tone
Concise. Technical. Cite the exact file(s) you used.
- Start with the answer, then the evidence.
- Prefer tables and bullets over prose.
- Distinguish paper claims from your own framing.
- Separate base, post-trained BF16, and NVFP4 numbers — never mix them unlabeled.
- Do not speculate beyond the sources.
What ships with it
19 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- context/index.toml 4.2 KB
- context/quick-reference.md 9.4 KB
- INDEX.md 12 KB
- model-card.md 6.3 KB
- paper/_overview.md 5.6 KB
- paper/architecture.md 4.1 KB
- paper/data.md 7.6 KB
- paper/evaluation.md 13 KB
- paper/inference.md 9.3 KB
- paper/infrastructure.md 11 KB
- paper/mopd/mtp-reasoning.md 8.6 KB
- paper/mopd/overview.md 8.3 KB
- paper/mopd/teachers.md 10 KB
- paper/mopd/warmup-results.md 8.9 KB
- paper/pretraining.md 8.3 KB
- paper/quantization.md 12 KB
- paper/safety.md 9.2 KB
- paper/sft.md 10 KB
- recipes/overview.md 3.2 KB
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 11d ago First seen · 165 lines · 68 tokens per session scan A 0343b9ad0f76
nemotron-ultra is a skill published in the GitHub repository NVIDIA-NeMo/Nemotron (2,043 stars, last pushed 4d ago), licensed Apache-2.0. It adds 68 tokens to every session and 1,867 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
tinker-fine-tuning
Provides guidance for fine-tuning LLMs using the Tinker cloud training API from Thinking Machines Lab. Use when running supervised fine-tuning, reinforcement learning (GRPO/PPO), or LoRA training on cloud GPUs via Tinker's managed infrastructure instead of local compute.
fine-tuning-with-trl
Fine-tune LLMs using reinforcement learning with TRL - SFT for instruction tuning, DPO for preference alignment, PPO/GRPO for reward optimization, and reward model training. Use when need RLHF, align model with preferences, or train from human feedback. Works with HuggingFace Transformers.
huawei-cloud-cloudrobo-train
Manage CloudRobo model training tasks and simulation reinforcement learning (SimRL) tasks — create pretrain (TRAINFROMSCRATCH) and finetune (MODELTUNING) tasks with FFT/SFT/LORA/QLORA/DEEPSPEED methods; manage the full task lifecycle (create/read/update/delete/stop/restart/resume/draft); save and resubmit draft…
fine-tuning-with-trl
Fine-tune LLMs using reinforcement learning with TRL - SFT for instruction tuning, DPO for preference alignment, PPO/GRPO for reward optimization, and reward model training. Use when need RLHF, align model with preferences, or train from human feedback. Works with HuggingFace Transformers.
fine-tuning-with-trl
Fine-tune LLMs using reinforcement learning with TRL - SFT for instruction tuning, DPO for preference alignment, PPO/GRPO for reward optimization, and reward model training. Use when need RLHF, align model with preferences, or train from human feedback. Works with HuggingFace Transformers.
fine-tuning-with-trl
Fine-tune LLMs using reinforcement learning with TRL - SFT for instruction tuning, DPO for preference alignment, PPO/GRPO for reward optimization, and reward model training. Use when need RLHF, align model with preferences, or train from human feedback. Works with HuggingFace Transformers.