What are the hardware requirements for Nemotron 3 Labs Ultra Math SFT?

Nemotron 3 Labs Ultra Math SFT requires approximately 85.9 GB of VRAM and supports a context window of 256K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Nemotron 3 Labs Ultra Math SFT and how large is it?

Nemotron 3 Labs Ultra Math SFT is developed by nvidia, a model with 550b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

Where can I download or evaluate Nemotron 3 Labs Ultra Math SFT?

Nemotron 3 Labs Ultra Math SFT is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

Nemotron 3 Labs Ultra Math SFT by nvidia — VRAM 85.9GB, 256K context

Name: Nemotron 3 Labs Ultra Math SFT
Author: nvidia

Nemotron 3 Labs Ultra Math SFT is an open-source language model by nvidia. Features: 550b LLM, VRAM: 85.9GB, Context: 256K, License: other, LLM Explorer Score: 0.32.

Arxiv:2609.10712 Base model:finetune:nvidia/nvi... Base model:nvidia/nvidia-nemot... Conversational Custom code Dataset:nvidia/nemotron-math-p... En Endpoints compatible Latent-moe Mtp Nemotron-3 Nemotron h Nvidia Pytorch Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/nvidia/Nemotron-3-Labs-Ultra-Math-SFT

Nemotron 3 Labs Ultra Math SFT Parameters and Internals

LLM Name	Nemotron 3 Labs Ultra Math SFT
Repository 🤗	https://huggingface.co/nvidia/Nemotron-3-Labs-Ultra-Math-SFT
Base Model(s)	...emotron 3 Ultra 550B A55B BF16 nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16
Model Size	550b
Required VRAM	85.9 GB
Updated	2026-09-11
Maintainer	nvidia
Model Type	nemotron_h
Model Files	2.1 GB: 1-of-529 2.1 GB: 2-of-529 2.1 GB: 3-of-529 2.1 GB: 4-of-529 2.1 GB: 5-of-529 2.1 GB: 6-of-529 2.1 GB: 7-of-529 2.1 GB: 8-of-529 2.1 GB: 9-of-529 2.1 GB: 10-of-529 2.1 GB: 11-of-529 1.9 GB: 12-of-529 2.1 GB: 13-of-529 2.1 GB: 14-of-529 2.1 GB: 15-of-529 2.1 GB: 16-of-529 2.1 GB: 17-of-529 2.1 GB: 18-of-529 2.1 GB: 19-of-529 2.1 GB: 20-of-529 2.1 GB: 21-of-529 2.1 GB: 22-of-529 2.1 GB: 23-of-529 2.1 GB: 24-of-529 2.1 GB: 25-of-529 2.1 GB: 26-of-529 2.1 GB: 27-of-529 2.1 GB: 28-of-529 2.1 GB: 29-of-529 2.1 GB: 30-of-529 2.1 GB: 31-of-529 2.1 GB: 32-of-529 2.1 GB: 33-of-529 2.1 GB: 34-of-529 2.1 GB: 35-of-529 2.1 GB: 36-of-529 2.1 GB: 37-of-529 2.1 GB: 38-of-529 2.1 GB: 39-of-529 2.1 GB: 40-of-529 2.1 GB: 41-of-529
Supported Languages	en
Model Architecture	NemotronHForCausalLM
License	other
Context Length	262144
Model Max Length	262144
Transformers Version	4.57.0
Vocabulary Size	131072

Rank the Nemotron 3 Labs Ultra Math SFT capabilities

Have you tried this model? Rate its performance — this feedback helps the ML community find the right model for their needs.

Instruction Following and Task Automation
Factuality and Completeness of Knowledge
Censorship and Alignment
Data Analysis and Insight Generation
Text Generation
Text Summarization and Feature Extraction
Code Generation
Multi-Language Support and Translation

Best Alternatives to Nemotron 3 Labs Ultra Math SFT

Best Alternatives	Context / RAM	Downloads	Likes
...motron 3 Ultra 550B A55B NVFP4	256K / 63.1 GB	240041	277
...emotron 3 Ultra 550B A55B BF16	256K / 204.9 GB	493863	308
...motron 3 Ultra 550B A55B NVFP4	256K / 63.1 GB	103	5
Nemotron 3 Labs Ultra Math RL	256K / 204.9 GB	374	3
...motron 3 Ultra 550B A55B GenRM	256K / 219.9 GB	2666	11
...on 3 Ultra 550B A55B Base BF16	256K / 209.9 GB	1005	29
Nemotron 3 Ultra 550B A55B	256K / 144.3 GB	473	2
...emotron 3 Ultra 550B A55B Base	256K / 209.9 GB	23	2
...DIA Nemotron 3 Ultra 550B A55B	256K / 204.9 GB	19	1
...B Abliterated Uncensored NVFP4	256K / 63.1 GB	582	5

Note: green Score (e.g. "73.2") means that the model is better than nvidia/Nemotron-3-Labs-Ultra-Math-SFT.