라이브
Head to head

Llama 3.1-405B vs Llama Nemotron Ultra 253B

Meta AI
Llama 3.1-405B
July 2024
vs
3.8×10²⁵Training compute (FLOP)3.9×10²⁵
$53MTraining cost--

Llama 3.1-405B (Meta AI) and Llama Nemotron Ultra 253B (NVIDIA) are both frontier AI models. Llama 3.1-405B was published in July 2024 and Llama Nemotron Ultra 253B in March 2025.

Llama Nemotron Ultra 253B was trained on 3.9×10²⁵ FLOP, essentially the same compute as Llama 3.1-405B at 3.8×10²⁵ FLOP. Training compute is the closest available proxy for how much was invested in a model, though it says nothing on its own about how well that compute was spent.

These two models share no benchmark on which both have been scored, so no direct performance comparison is possible here. The specification table below is a comparison of inputs, not of results.

Specifications
Meta AI
Organization
NVIDIA
Jul 23, 2024
Published
Mar 18, 2025
3.8×10²⁵ FLOP
Training compute1.0x
3.9×10²⁵ FLOP
405B
Parameters1.6x
253B
15.6T
Dataset size26x
603B
NVIDIA H100 SXM5 80GB
Training hardware
--
16,384
Chips used
--
2,142 h
Training time
--
$53M
Training cost (2023 USD)
--
22.6 MW
Training power draw
--
Open weights (restricted use)
Accessibility
Open weights (restricted use)
Yes
Open weights
Yes
United States
Country
United States
Related comparisons
SourceEpoch AI, 'AI Models'. Published online at epoch.ai. Retrieved 2026-07-29 from https://epoch.ai/data/ai-models. Licensed under CC BY 4.0.