Nemotron 3 Ultra (550B A55B)

Name: Nemotron 3 Ultra (550B A55B)
Author: NVIDIA

NVIDIA

Nemotron 3 Ultra is NVIDIA's frontier-scale open model with 550B total / 55B active parameters, built for agentic reasoning, long-context analysis, tool use, and high-stakes RAG.

Key Specifications

Parameters

550.0B

Context

Release Date

June 4, 2026

Average Score

64.8%

API Documentation Research Paper Repository Model Weights Results Blog

Timeline

Key dates in the model's history

Announcement

June 4, 2026

Last Update

August 29, 2026

Today

October 17, 2026

Technical Specifications

Parameters

550.0B

Training Tokens

20.0T tokens

Knowledge Cutoff

September 30, 2025

Family

Capabilities

MultimodalZeroEval

Benchmark Results

Model performance metrics across various tests and benchmarks

Programming

Programming skills tests

SWE-Bench Verified

• Self-reported

70.7%

Reasoning

Logical reasoning and analysis

GPQA

No tools • Self-reported

87.0%

Other Tests

Specialized benchmarks

AA-LCR

• Self-reported

65.4%

Apex

Shortlist (with tools) • Self-reported

84.8%

BrowseComp

With search • Self-reported

44.4%

CritPT

No tools • Self-reported

3.1%

Finance Agent

Vals.ai Financial Agent 1.1 (with web search) • Self-reported

53.7%

Finance Agent v2

• Verified

37.5%

GDPval

• Self-reported

46.7%

Humanity's Last Exam

With tools • Self-reported

37.4%

IFBench

Prompt loose • Self-reported

81.7%

IMO-AnswerBench

With tools • Self-reported

92.3%

LiveCodeBench v6

• Self-reported

89.0%

LongBench v2

≤ 1M • Self-reported

61.9%

MMLU-Pro

• Self-reported

86.8%

MMLU-ProX

Average over languages • Self-reported

83.0%

Multi-Challenge

• Self-reported

63.8%

OmniScience

Non-Hallucination • Self-reported

78.7%

PinchBench

• Self-reported

90.0%

ProfBench

Search • Self-reported

56.0%

RULER

1M • Self-reported

94.7%

SciCode

Subtask • Self-reported

44.6%

SWE-bench Multilingual

• Self-reported

67.7%

TAU3-Bench

Banking • Self-reported

22.6%

Terminal-Bench 2.1

• Self-reported

56.4%

WMT24++

en→xx • Self-reported

83.7%

License & Metadata

License

openmdw_license_v1_1

Announcement Date

June 4, 2026

Last Updated

August 29, 2026

Similar Models

All Models

Nemotron 3 Super (120B A12B)

NVIDIA

120.0B

Best score:0.8 (GPQA)

Released:Mar 2026

Llama 3.1 Nemotron Ultra 253B v1

NVIDIA

253.0B

Best score:0.8 (GPQA)

Released:Apr 2025

LongCat-Flash-Thinking-2601

Meituan

560.0B

Best score:1.0 (TAU)

Released:Jan 2026

GLM-5.1

Zhipu AI

754.0B

Best score:0.9 (GPQA)

Released:Apr 2026

Price:$1.40/1M tokens

GLM-4.7

Zhipu AI

358.0B

Best score:0.9 (TAU)

Released:Dec 2025

Price:$0.60/1M tokens

MiMo-V2-Flash

Xiaomi

309.0B

Best score:0.8 (GPQA)

Released:Dec 2025

Hy4 preview

Tencent

770.0B

Best score:0.9 (GPQA)

Released:Aug 2026

Hy3

Tencent

295.0B

Best score:0.9 (GPQA)

Released:Jul 2026

Recommendations are based on similarity of characteristics: developer organization, multimodality, parameter size, and benchmark performance. Choose a model to compare or go to the full catalog to browse all available AI models.