2.3× less energy than Nemotron 3 Super.
Nemotron 3 Super Environmental Impact
Leading open-weight model — 120B total / 12B active, 2.2x faster than GPT-OSS
One query, in context
Energy, carbon and water for a single use. Short query (~300 tokens)
About 1.5× a Google search and 54% of a minute of laptop use.
About 1.7× a Google search and 10% of 10 m by car.
About 32× a drop of water and 32% of a teaspoon of water.
- Where it runsSelf-hosted / NVIDIA NIM
- Grid carbon intensity458 g CO₂/kWh · 32% renewable
- Compared with a Google search2x more than a search (0.3 Wh)
- Water~625 queries to fill 1 litre
Where Nemotron 3 Super sits
#108 of 199 by energy per query. Heavier than 53% of all models. Lit dots are the same kind of model; click any dot to open it.
Want something lighter?
1.9× less energy than Nemotron 3 Super.
6.4× less energy than Nemotron 3 Super.
What a month of use adds up to
Pick how much you use it. Based on one quick use at a time, five days a week.
- Energy
- 48.7 Wh
- Carbon
- 21.7 g
- Water
- 173 mL
A month of 5 chats a day, 5 days a week. The energy is about 4.1× a phone charge and 44% of a kettle boil.
Work out your own →- Energy
- 194.8 Wh
- Carbon
- 86.6 g
- Water
- 693 mL
A month of 20 chats a day, 5 days a week. The energy is about 1.8× a kettle boil and 3% of a UK home's day.
Work out your own →- Energy
- 974.3 Wh
- Carbon
- 433.0 g
- Water
- 3.46 L
A month of 100 chats a day, 5 days a week. The energy is about 8.9× a kettle boil and 13% of a UK home's day.
Work out your own →How does Nemotron 3 Super compare?
Next to similar models. Bars use a log scale, so each step is a big jump.
Detailed breakdown
Energy
Nemotron 3 Super is NVIDIA's flagship open-weight model at 120B total / 12B active parameters (MoE). At ~0.45 Wh per query, it is 2.2x faster than GPT-OSS 120B while matching or exceeding its quality. The sparse architecture activates only 10% of parameters per token, making it extremely efficient. Optimised for NVIDIA hardware via NIM (NVIDIA Inference Microservices).
Power source & carbon
Open-weight. Optimised for NVIDIA GPUs via TensorRT-LLM. Available through NVIDIA NIM and popular on OpenRouter.
Water use
At ~1.6 mL per query — remarkably low for a model with 120B total parameters thanks to MoE efficiency.
About Nemotron 3 Super
Nemotron 3 Super is a 120B-parameter text and chat model from NVIDIA, released March 11, 2026. Leading open-weight model — 120B total / 12B active, 2.2x faster than GPT-OSS. At 0.45 Wh per query, it uses roughly the same energy as a Google search. It runs on a Mixture-of-Experts (12B active) architecture.
These figures are estimates derived from hardware specifications and API benchmarks — NVIDIA has not published official energy data for Nemotron 3 Super. Actual consumption may vary significantly depending on batching, quantisation, and infrastructure optimisations that we cannot observe from outside.
Nemotron 3 Super in context
Key insights
- Open-source weights — can be self-hosted on infrastructure you control
What does your Nemotron 3 Super usage cost the planet?
Use the calculator to estimate your personal footprint based on how often you use Nemotron 3 Super.
Frequently asked questions
How much energy does Nemotron 3 Super use per query?
Each Nemotron 3 Super query consumes approximately 0.45 Wh of energy. This is 2x more than a traditional Google search (~0.3 Wh).
What is Nemotron 3 Super's carbon footprint?
Based on the carbon intensity of Self-hosted / NVIDIA NIM, each query produces approximately 0.20 g of CO2. The grid in this region has a carbon intensity of 458 g CO2/kWh with 32% renewable energy.
How much water does Nemotron 3 Super use?
Each query consumes approximately 2 mL of water, primarily used for cooling the data centers that process the request.
How does Nemotron 3 Super compare to a Google search?
A Nemotron 3 Super query uses 2x more than a Google search in terms of energy. A Google search uses approximately 0.3 Wh, while Nemotron 3 Super uses 0.45 Wh.
Technical details
- Architecture
- Mixture-of-Experts (12B active)
- Parameters
- 120B
- Context window
- 128,000 tokens
- Release date
- 2026-03-11
- Open source
- Yes
- Training data cutoff
- 2026-02