RunLocalAI
llama
253B parameters
Commercial OK

Llama 3.1 Nemotron Ultra 253B

NVIDIA's top open reasoning model in the Llama 3.1 lineage. Server-tier; trained for groundbreaking reasoning accuracy on agentic workloads.

License: NVIDIA Open Model License·Released Apr 8, 2025·Context: 131,072 tokens

Overview

NVIDIA's top open reasoning model in the Llama 3.1 lineage. Server-tier; trained for groundbreaking reasoning accuracy on agentic workloads.

Strengths

  • Top open reasoning at release
  • Optimized for NVIDIA hardware

Weaknesses

  • Server-only
  • 160GB+ VRAM

Quantization variants

Each quantization trades model quality for file size and VRAM. Q4_K_M is the most popular starting point.

QuantizationFile sizeVRAM required
Q4_K_M144.0 GB160 GB

Get the model

HuggingFace

Original weights

huggingface.co/nvidia/Llama-3_1-Nemotron-Ultra-253B-v1

Source repository — direct quantization required.

Hardware that runs this

Cards with enough VRAM for at least one quantization of Llama 3.1 Nemotron Ultra 253B.

NVIDIA GB200 NVL72
13824GB · nvidia
AMD Instinct MI355X
288GB · amd
AMD Instinct MI325X
256GB · amd
NVIDIA B200
192GB · nvidia
AMD Instinct MI300X
192GB · amd
NVIDIA H100 NVL
188GB · nvidia
Compare alternatives

Models worth comparing

Same parameter band, plus what's one tier above and below — so you can decide what actually fits your hardware.

Step up
More capable — bigger memory footprint
No verdicted models in the next tier up yet.

Frequently asked

What's the minimum VRAM to run Llama 3.1 Nemotron Ultra 253B?

160GB of VRAM is enough to run Llama 3.1 Nemotron Ultra 253B at the Q4_K_M quantization (file size 144.0 GB). Higher-quality quantizations need more.

Can I use Llama 3.1 Nemotron Ultra 253B commercially?

Yes — Llama 3.1 Nemotron Ultra 253B ships under the NVIDIA Open Model License, which permits commercial use. Always read the license text before deployment.

What's the context length of Llama 3.1 Nemotron Ultra 253B?

Llama 3.1 Nemotron Ultra 253B supports a context window of 131,072 tokens (about 131K).

Source: huggingface.co/nvidia/Llama-3_1-Nemotron-Ultra-253B-v1

Reviewed by RunLocalAI Editorial. See our editorial policy for how we research and verify model claims.