NVIDIA

Nemotron 3 Super

by NVIDIAtext->text2 endpoints
Neura Intelligence Index
45.6/ 100
Rank
#198
Confidence
medium
Benchmarks
7(18% coverage)

Overview

NVIDIA Nemotron 3 Super is a 120B-parameter open hybrid MoE model, activating just 12B parameters for maximum compute efficiency and accuracy in complex multi-agent applications. Built on a hybrid Mamba-Transformer...

Capabilities

Text generation
Tool use / Function calling
Structured output
Extended reasoning
Long context
Top-P sampling
Stop sequences
Deterministic seed
Log probabilities

Modalities

Input
Text
Output
Text

Technical Specifications

Context Window
262.1K tokens
Max Output
235.9K tokens
Tokenizer
Other
Uptime (24h)
9999.3%
Parameters
120000000000B
License
nvidia_open_model_license_agreementOpen Weights
Country
🇺🇸US
Release Date
March 11, 20266mo

Supported Parameters

frequency_penaltyinclude_reasoninglogit_biaslogprobsmax_tokensmin_ppresence_penaltyreasoningreasoning_effortrepetition_penaltyresponse_formatseedstopstructured_outputstemperaturetool_choicetoolstop_ktop_logprobstop_p

Benchmark Performance

coding

BenchmarkScoreSource
SWE-bench Verified
53.7
verified
SciCode
42.0
verified
TerminalBench
25.8
verified

math

BenchmarkScoreSource
AIME 2025
90.2
verified

reasoning

BenchmarkScoreSource
GPQA Diamond
82.7
verified
Humanity's Last Exam
22.8
verified

agent

BenchmarkScoreSource
BrowseComp
31.3
verified

Performance Over Time

Pricing

Per 1M tokens
Input$0.08
Output$0.45
Added
March 11, 2026
Last synced 9/21/2026