modelbenchmark.io

Nemotron 3 Ultra

NVIDIA · nvidia-nemotron-3-ultra-550b-a55b

Compare

Nemotron 3 Ultra is a powerful MoE model designed for long-running agents across coding, deep research, and enterprise automation.

Specification

most-agreed values

Context
203K
Max output
203K
Released
2026-06-04
Knowledge cutoff
2025-09
Retires
—
Open weights
yes
Input
text
Output
text

Price

US dollars per million tokens · most-agreed

Input
$0.60
Output
$2.40
Cache read
$0.12

Available from 5 hosts

HostIn $/MOut $/MCache rdCache wrContextOutputRetires
deepinfra$0.50$2.20$0.10—262K262K—
wandb$0.50$2.15$0.10—262K262K—
wandb · wandb$0.50$2.15$0.10—262K262K—
baseten$0.60$2.40$0.12—203K203K—
venice$0.625$3.125$0.1875—256K33K—

Price history

append-only observations · a listing writes a row only when its price moves

DateHostField$/M
2026-09-02basetenCache read$0.12
2026-09-23basetenCache read$0.12
2026-09-02wandbCache read$0.15
2026-09-15wandbCache read$0.10
2026-09-02wandb · wandbCache read$0.15
2026-09-18wandb · wandbCache read$0.10
2026-09-02basetenInput$0.60
2026-09-23basetenInput$0.60
2026-09-02wandbInput$0.75
2026-09-15wandbInput$0.50
2026-09-02wandb · wandbInput$0.75
2026-09-18wandb · wandbInput$0.50
2026-09-02basetenOutput$2.40
2026-09-23basetenOutput$2.40
2026-09-02wandbOutput$2.75
2026-09-15wandbOutput$2.15
2026-09-02wandb · wandbOutput$2.75
2026-09-18wandb · wandbOutput$2.15

More from NVIDIA

most-hosted first