Back to Explore
NVIDIA: Nemotron 3 Ultra
NVIDIA·Released 2026-06-04
Open Source1.0M ctx
Comparison data ready55% coverage10/10 fields directly observedUpdated Jul 19, 2026, 5:30 PM
About
NVIDIA Nemotron 3 Ultra is an open frontier-reasoning and orchestration model from NVIDIA, with 55B active parameters out of 550B total (MoE). Built on a hybrid Transformer-Mamba mixture-of-experts architecture, it...
Related Models
Nemotron 3 Ultra 550B A55B (Reasoning)2026-06-04NVIDIA: Nemotron 3 Ultra (free)2026-06-04NVIDIA: Nemotron 3.5 Content Safety (free)2026-06-04Nemotron 3 Nano Omni 30B A3B Reasoning2026-04-29NVIDIA: Nemotron 3 Nano Omni (free)2026-04-28Nemotron Cascade 2 30B A3B2026-03-19NVIDIA: Nemotron 3 Super2026-03-11NVIDIA: Nemotron 3 Super (free)2026-03-11
Pricing
Input
$0.60
per 1M tokens
Output
$3.60
per 1M tokens
Blended
$1.35
per 1M tokens
Cheaper than 35% of models. Median price is $0.70/1M tokens.
Cost Calculator
Tokens per day1M
100K100M
Daily
$1.35
Monthly
$40.50
vs. Similar Models
Qwen3.6 27B (Reasoning)
$1.35+0%
Qwen3.5 397B A17B (Reasoning)
$1.35+0%
Qwen3.5 397B A17B (Non-reasoning)
$1.35+0%
Qwen3.6 27B (Non-reasoning)
$1.35+0%
Performance
Context Window
1.0M
tokens
Larger than 79% of models
Max Output
16K
tokens
2% of context
Context Window Comparison
Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback)
1.0MSame
Claude Opus 4.8 (Adaptive Reasoning, Max Effort)
1.0MSame
Claude Opus 4.7 (Adaptive Reasoning, Max Effort)
1.0MSame