models/nvidia/nemotron-3-super-120b-a12b
N
NVIDIA·active·free tier available

NVIDIA: Nemotron 3 Super

NVIDIA's mid-tier model. Long-context specialist with 1M window.

Overall score
4.46
/5.00 · ranked #49
Input
$0.085
per 1M tokens
Output
$0.400
per 1M tokens
Context
1M
tokens
Blended
$0.321
3:1 out:in ratio

Price drops, new benchmarks, model updates. Stay current on NVIDIA: Nemotron 3 Super.

One email per change. Unsubscribe anytime.

modelpicker.aipowered by live benchmark data

Scores by test

Methodology →
Structured Output
5.0
Strategic Analysis
5.0
Constrained Rewriting
5.0
Creative Problem Solving
4.0
Tool Calling
4.0
Faithfulness
5.0
Classification
4.0
Long Context
5.0
Safety Calibration
2.0
Persona Consistency
4.0
Agentic Planning
5.0
Multilingual
5.0
Tabular Data
5.0

What you need to know

NVIDIA Nemotron 3 Super is a high-performance model optimized for complex structural and analytical tasks. It achieves perfect internal scores in long context handling, structured output, strategic analysis, and agentic planning. With a 262K context window and a 5/5 rating for faithfulness and tabular data, it is specifically engineered for high-precision data extraction and reasoning over large datasets.

At a blended cost of $0.360 per million tokens, the model offers a high value-to-performance ratio. It ranks 16th out of 71 models overall, providing top-tier analytical capabilities at a price point significantly lower than many proprietary frontier models. This makes it a cost-effective choice for enterprise-grade automation that requires strict adherence to formats and strategic depth.

The model's primary limitation is safety calibration, where it scores 2/5, indicating a potential lack of robust guardrails or a tendency to bypass safety constraints. It also shows slight relative weaknesses in creative problem solving and tool calling compared to its perfect scores in analysis and structure.

Use this model if your workflow requires processing massive documents, generating strict JSON or tabular outputs, or executing complex agentic planning. Skip this model if your application requires high safety calibration or if the primary goal is open-weight deployment.

Strengths — Top 3

Structured Output5.0/5.0
Strategic Analysis5.0/5.0
Constrained Rewriting5.0/5.0

Relative weaknesses — Bottom 3

Safety Calibration2.0/5.0
Creative Problem Solving4.0/5.0
Tool Calling4.0/5.0

Similar models

QQwen: Qwen3.6 Plus$1.544.54QQwen: Qwen3.7 Flash$0.1054.54OGPT-5$7.814.54DDeepSeek V3.2$0.3674.31