Compare AI models

Specifications
Parameters
16B550B
Context window
32k1M
API pricing
Cheapest input
—$0.40DeepInfra
Cheapest output
—$1.76DeepInfra
—$210KT
These models have no shared benchmark scores.
Overview
CompanyDeepSeekNVIDIA
Release dateMay 16 2024Jun 4 2026
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

DeepSeek V2 Lite and Nemotron 3 Ultra don't publish scores on any of the same benchmarks, so there's no direct head-to-head comparison. DeepSeek V2 Lite shipped 749 days before Nemotron 3 Ultra, so benchmark comparisons should account for the intervening progress.

DeepSeek V2 Lite has 16B parameters, while Nemotron 3 Ultra has 550B. Context windows are 32k (DeepSeek V2 Lite) vs 1M (Nemotron 3 Ultra).

Direct benchmark comparisons are unavailable — DeepSeek V2 Lite and Nemotron 3 Ultra don't publish scores on any of the same benchmarks.