Nemotron 3 Super
Nemotron 3 Super
Open weights tuned by the people who make the chips, and it decodes accordingly.
Updated
The read
Nvidia builds Nemotron to show what its own hardware can do, which is why this model is optimised for throughput to a degree nobody else bothers with. The weights are open under Nvidia's commercially usable licence, quality is solid across general and coding work, and if you are already serving on their stack it will be the fastest thing you can run. It is not trying to win a reasoning benchmark.
Where it shines
- Among the fastest open models measured, by a wide margin
- Licence permits commercial use without a revenue threshold
- Tuned in lockstep with the inference stack most people deploy on
Worth knowing first
- Optimised for Nvidia hardware, so gains shrink elsewhere
- Mid-pack on reasoning against open rivals of similar size
Index score
80.1
Provisional. This score is read from the product's public capability, not from the one-prompt rebuild the fully reviewed entries went through, so treat it as a placing rather than a verdict.
- Rank
- 24 of 53
- Builds
- LLMs
- Category
- Open weights
- Output
- Open weights under Nvidia's commercial licence
- Pricing
- Free tier
- Best for
- Maximum throughput on hardware you already run
Scores are our own editorial judgement, weighted craft 55, speed 10, control 15 and value 20.
Also worth a look
Full index- 05Kimi K3Kimi K3ProvisionalThe largest open-weights model anyone has shipped, and it reads a million tokens without blinking.Open weightsFree tier87.5
- 06GLM-5.2GLM-5.2ProvisionalFrontier-class output, an MIT licence, and a million tokens of context. That combination is the story.Open weightsFree tier87.1
- 12Qwen 3.6Qwen 3.6ProvisionalThe open Qwen release most people should actually run, in sizes that fit real hardware.Open weightsFree tier83.3
