Nemotron 3 Ultra

Nemotron 3 Ultra is a reasoning model from NVIDIA. 24 benchmarks count toward its index, in 6 categories.

  • Open weights
  • Reasoning
  • Input: text
  • Released 2026-06
RankedEnough results for an index.
IndexAll categories, out of 100.50.8±5.3Median 51.0
CoverageIndex weight that has results.80%
SpeedOutput tokens per second.—Median 67/s
First tokenSeconds to the first output token.—Median 1.3 s
Input / 1MUS dollars per 1M input tokens.FreeMedian $0.60
Output / 1MUS dollars per 1M output tokens.FreeMedian $2.55
ContextMaximum tokens in one request.1MMedian 262K
EloLMArena rating and rank.—Median 1422

Run locally

560.52B parameters
16-bit
1178.1 GB
8-bit
626.3 GB
4-bit
357.8 GB
Weights
Hugging Face (opens in a new tab) ↗

Memory need with a short context. The 4-bit copy needs more than 128 GB. Compare local models

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.12 counted48.9135 / 218
CodingCode writing and repair.3 counted54.3105 / 209
ReasoningLogic problems and puzzles.3 counted50.0116 / 227
MultimodalTasks with images and text.None counted——
KnowledgeFacts and expert knowledge.4 counted50.0135 / 268
MultilingualTasks in many languages.1 counted52.4—
InstructionTasks with strict prompt rules.1 counted65.2—
MathMath problems.None counted——

Results

24 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
Artificial Analysis IFBenchInstruction81.4%2 / 146
Instruction Following BenchmarkInstruction81.7%5 / 57
LiveCodeBench v6Coding89.0%12 / 59
PinchBenchAgentic90.0%3 / 10
Artificial Analysis GPQA DiamondKnowledge86.7%62 / 199
Graduate-Level Google-Proof Q&AKnowledge87.0%44 / 141
LongBench v2Reasoning61.9%9 / 25
MMLU-ProXMultilingual83.0%7 / 17
GDPval-AA normalizedAgentic33.1%55 / 133
Scientific Code BenchmarkCoding44.6%15 / 35
Critical Physics TasksReasoning3.1%95 / 214
GPQA Diamond · BenchLMKnowledge87.0%49 / 84
Humanity's Last ExamKnowledge26.7%55 / 77
Artificial Analysis SciCodeCoding40.3%86 / 114
BrowseCompAgentic44.4%61 / 63

Sources

BenchLMCC BY-NC 4.0

History

5 releases
Index53.2 → 50.8
23 Oct 202612 Nov 2026
Output / 1MFree → Free
23 Oct 202612 Nov 2026

Questions

How much does Nemotron 3 Ultra cost?
Input tokens are free, output tokens are free.
How good is Nemotron 3 Ultra?
Its index is 50.8 out of 100. That is place 145 of 283 ranked models.
What is Nemotron 3 Ultra best at?
Coding: 54.3 out of 100, place 105 of 209.
What is the context window of Nemotron 3 Ultra?
1M tokens.
Is Nemotron 3 Ultra open source?
Yes. Its weights are published.
When was Nemotron 3 Ultra released?
June 2026.

Badge

Nemotron 3 Ultra on Inference360: place 145 of 283, index 50.8
HTML
<a href="https://inference360.ai/models/nemotron-3-ultra"><img src="https://inference360.ai/badges/nemotron-3-ultra.svg" alt="Nemotron 3 Ultra on Inference360: place 145 of 283, index 50.8" height="20"></a>
Markdown
[![Nemotron 3 Ultra on Inference360: place 145 of 283, index 50.8](https://inference360.ai/badges/nemotron-3-ultra.svg)](https://inference360.ai/models/nemotron-3-ultra)

Compare with

Nemotron 3.5 Super VL 120B A12B NVFP450.8Ling 3.0 Flash50.9Qwen3.6-35B-A3B50.6Mistral Medium 3.5 128B50.3Nemotron 3 Nano Omni 30B A3B38.6

More from NVIDIA

All models
Nemotron 3.5 ASR Streaming 0.6BUnrankedNemotron 3.5 Content SafetyUnrankedCosmos3-NanoUnrankedCosmos3-SuperUnrankedNemotron 3 Nano Omni 30B A3B38.6