Compare AI models

Specifications
Parameters
671B550B
Context window
128k1M
API pricing
Cheapest input
—$0.40DeepInfra
Cheapest output
—$1.76DeepInfra
—$210KT
Benchmarks
SWE-Bench Verified
49.2%71.9%
GPQA Diamond
71.5%86.7%
Overview
CompanyDeepSeekNVIDIA
Release dateJan 20 2025Jun 4 2026
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

Nemotron 3 Ultra leads DeepSeek R1 on 2 of the 2 benchmarks they both report (SWE-Bench Verified, GPQA Diamond). DeepSeek R1 shipped 500 days before Nemotron 3 Ultra, so benchmark comparisons should account for the intervening progress.

DeepSeek R1 has 671B parameters, while Nemotron 3 Ultra has 550B. Context windows are 128k (DeepSeek R1) vs 1M (Nemotron 3 Ultra).

On SWE-Bench Verified, Nemotron 3 Ultra leads at 71.9% vs DeepSeek R1 at 49.2%. On GPQA Diamond, Nemotron 3 Ultra leads at 86.7% vs DeepSeek R1 at 71.5%.