Nemotron 3 Ultra

Nemotron 3 Ultra is a reasoning model from NVIDIA. 25 benchmarks count toward its score, in 5 categories.

availableShows if the model has enough results for an index.
IndexOverall score out of 100.53.4 ±5.7
CoverageShare of the index weight with results.75%
SpeedOutput tokens per second.
Input / 1MUS dollars per 1M input tokens.Free
Output / 1MUS dollars per 1M output tokens.Free
ContextMaximum tokens in one request.1M
EloLMArena rating and rank.N/A

The index is a score out of 100. The ± range shows how much it can change. A free tier is available.

CapabilitiesScore per category, out of 100.

Out of 100
AgenticMulti-step tasks with tools.
52.5
CodingCode writing and repair.
54.1
ReasoningLogic problems and puzzles.
49.7
MultimodalTasks with images and text.
N/A
KnowledgeFacts and expert knowledge.
53.1
MultilingualTasks in many languages.
N/A
InstructionTasks with strict rules in the prompt.
66.4
MathMath problems.
N/A

Results

25 counted
BenchmarkThe test name.CategoryThe capability that the test measures.ResultThe score from the publisher.IndexThis result as a score out of 100.RunThe settings of the run.DateDate of the result.Published byThe source of the result.
PinchBenchAgentic90.0%Kilo Code
LiveCodeBench v6Coding89.0%58.0LiveCodeBench maintainers
Graduate-Level Google-Proof Q&AKnowledge87.0%58.6David Rein et al.
GPQA DiamondKnowledge87.0%58.6David Rein et al.
Massive Multitask Language Understanding ProfessionalKnowledge86.8%57.3Yubo Wang et al.
Artificial Analysis GPQA DiamondKnowledge86.7%57.4Artificial Analysis
τ²-Bench Tool-Agent-User EvaluationAgentic83.3%58.6Victor Barres et al.
Artificial Analysis Openness IndexKnowledge83.3%Artificial Analysis
MMLU-ProXMultilingual83.0%MMLU-ProX authors
Artificial Analysis Harvey LAB-AAAgentic81.7%61.2Artificial Analysis
Instruction Following BenchmarkInstruction81.7%61.1Benchmark authors
Artificial Analysis IFBenchInstruction81.4%71.7Artificial Analysis
Software Engineering Benchmark VerifiedCoding71.9%55.2Carlos E. Jimenez et al.
τ³-Bench Tool-Agent-User EvaluationAgentic70.9%55.4Sierra Research
Artificial Analysis Long Context ReasoningReasoning67.0%54.6Artificial Analysis
LongBench v2Reasoning61.9%LongBench v2 authors
Artificial Analysis Coding IndexCoding49.3%53.7Artificial Analysis
Scientific Code BenchmarkCoding44.6%54.8Benchmark authors
BrowseCompAgentic44.4%37.3OpenAI
Artificial Analysis SciCodeCoding40.3%48.8Artificial Analysis
Humanity's Last Exam with toolsAgentic37.4%50.9DeepSeek-AI
Terminal-Bench HardAgentic36.4%Artificial Analysis
GDPval-AA normalizedAgentic33.1%61.0Artificial Analysis
Artificial Analysis EnterpriseOps-GymAgentic28.9%47.0Artificial Analysis
Artificial Analysis Humanity's Last ExamKnowledge28.4%55.4Artificial Analysis
Humanity's Last ExamKnowledge26.7%51.4Center for AI Safety et al.
Humanity's Last Exam without toolsKnowledge26.7%51.4OpenAI
Artificial Analysis Intelligence IndexKnowledge22.9%51.1Artificial Analysis
Artificial Analysis Agentic IndexAgentic21.7%54.8Artificial Analysis
Artificial Analysis Omniscience AccuracyKnowledge21.6%40.6Artificial Analysis
Artificial Analysis AnalystAgentAgentic6.3%45.9Artificial Analysis
Critical Physics TasksReasoning3.1%44.9Artificial Analysis

25 benchmarks count, from 27 of 32 results. A grey row does not count. Too few models took that benchmark.

Sources

BenchLM benchmark aggregationUsed with attribution; per-benchmark results credited to their original authors

More from NVIDIA

Nemotron 3 Super 100B42.5Nemotron 3 Nano Omni 30B A3B36.9Nemotron 3.5 Lightning 30B A3B NVFP437.4Nemotron 3 Nano 30B37.1Audio Flamingo 3 7BUnranked