VENICE ATLAS · MODEL DOSSIER · NVIDIA NEMOTRON 3 ULTRA

Model overview

NVIDIA Nemotron 3 Ultra

Name: NVIDIA Nemotron 3 Ultra
Brand: venice.ai
Price: 0.6250 USD
Availability: InStock

NVIDIA Nemotron 3 Ultra is built for frontier reasoning, orchestration, coding agents, deep research, and complex enterprise workflows. It delivers up to 5x faster inference and up to 30% lower cost for agentic workloads while supporting up to 1M token context.

nvidia-nemotron-3-ultra-550b-a55b · venice.ai

Text · Venice catalog$0.63 per 1M tokensPrice on Venice AI

ModalityTextOn Venice AI

Primary price$0.63 per 1M tokensper 1M tokens

Context window256KMax tokens

Benchmarks26Benchmark results

Benchmark results

See the tests behind each score. They use the benchmark’s original scoring scale.

Biology#3329.72 rating Chat#724.36 rating Communication#1016.74 rating Finance#1628.84 rating Healthcare#834.49 rating Instruction Following#429.29 rating Knowledge#221.49 rating Language#534.36 rating Legal#533.49 rating Long Context#1421.48 rating Math#3036.00 rating Science#816.80 rating Structured Output#1122.81 rating

Agents3 results

Finance Agentfinance-agent · agents

53.7%

ReportedOct 2026

Finance Agent v2finance-agent-v2 · agents

37.5%

ReportedOct 2026

TAU3-Benchtau3-bench · agents

22.6%

ReportedOct 2026

Chat1 result

Multi-Challengemultichallenge · chat

63.8%

ReportedOct 2026

Coding1 result

PinchBenchpinchbench · code

90.0%

ReportedOct 2026

General intelligence1 result

GDPvalgdpval · general

46.7%

ReportedAug 2026

Instruction Following1 result

IFBenchifbench · instruction_following

81.7%

ReportedOct 2026

Language3 results

MMLU-Prommlu-pro · language

86.8%

ReportedOct 2026

MMLU-ProXmmlu-prox · language

83.0%

ReportedOct 2026

WMT24++wmt24++ · language

83.7%

ReportedOct 2026

Long Context2 results

LongBench v2longbench-v2 · long_context

61.9%

ReportedOct 2026

RULERruler · long_context

94.7%

ReportedOct 2026

Math1 result

IMO-AnswerBenchimo-answerbench · math

92.3%

ReportedOct 2026

Reasoning12 results

AA-LCRaa-lcr · reasoning

65.4%

ReportedOct 2026

Apexapex · reasoning

84.8%

ReportedOct 2026

BrowseCompbrowsecomp · reasoning

44.4%

ReportedOct 2026

CritPTcritpt · reasoning

3.10%

ReportedOct 2026

GPQAgpqa · reasoning

87.0%

ReportedOct 2026

Humanity's Last Examhumanity's-last-exam · reasoning

37.4%

ReportedOct 2026

LiveCodeBench v6livecodebench-v6 · reasoning

89.0%

ReportedOct 2026

ProfBenchprofbench · reasoning

56.0%

ReportedOct 2026

SciCodescicode · reasoning

44.6%

ReportedOct 2026

SWE-bench Multilingualswe-bench-multilingual · reasoning

67.7%

ReportedOct 2026

SWE-Bench Verifiedswe-bench-verified · reasoning

70.7%

ReportedOct 2026

Terminal-Bench 2.1terminal-bench-2.1 · reasoning

56.4%

ReportedOct 2026

Science1 result

OmniScienceomniscience · science

78.7%

ReportedOct 2026

Learn what each benchmark measures ↗

Pricing detail

Input price per 1M tokens: $0.63
Output price per 1M tokens: $3.13
Cache read price per 1M tokens: $0.19
Blended price per 1M tokens (50/50): $1.88

Prices are published by Venice AI. Venice model page ↗

Capabilities

ReasoningFunction callingWeb search

Endpoint metadata

Privacy: private
Quantization: fp8
Maximum completion tokens: 33K
Uncensored: No
Beta: No
Mapping confidence: 100.0%

View model source ↗

All capabilities

Quantization: fp8
Supports E2 E E: No
Supports Vision: No
Supports X Search: No
Optimized For Code: No
Supports Log Probs: No
Supports Reasoning: Yes
Supports Web Search: Yes
Supports Audio Input: No
Supports Video Input: No
Default Reasoning Effort: low
Reasoning Effort Options: none, low, medium, high
Supports Multiple Images: No
Supports Response Schema: Yes
Supports Tee Attestation: No
Supports Function Calling: Yes
Supports Reasoning Effort: Yes

Related models

More from nvidia

nvidia-nemotron-3-5-super-vl-120b-a12bNVIDIA Nemotron 3.5 Super VL 120B$0.25 input per 1M tokens ↗nvidia-nemotron-3-nano-30b-a3bNVIDIA Nemotron 3 Nano 30B$0.07 input per 1M tokens ↗