NVIDIA: Nemotron 3 Ultra pricing and limits

id nvidia/nemotron-3-ultra-550b-a55b · listed 2026-06-04

Input (per 1M tokens)	$0.600
Output (per 1M tokens)	$2.400
Cached input read (per 1M)	$0.120
Context window	262,144 tokens
Max output	182,520 tokens
Modalities	text → text
Reasoning	yes

What 1,000 calls cost at 1,000 input and 300 output tokens each: $1.32.

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...

Cheaper on both input and output, with at least half the context

Model	In / 1M	Out / 1M	Context
Auto Router (Beta)	$-1000000.000	$-1000000.000	2,000,000
Pareto Code Router	$-1000000.000	$-1000000.000	2,000,000
Auto Router	$-1000000.000	$-1000000.000	2,000,000
Meta: Llama 4 Scout	$0.100	$0.300	1,310,720
OpenAI: GPT-6 Luna Pro	$0.100	$0.500	1,050,000
OpenAI: GPT-6 Luna Pro (batch)	$0.050	$0.250	1,050,000

Other nvidia models

NVIDIA: Switchyard ($-1000000.000 / $-1000000.000) · NVIDIA: Nemotron 3.5 Super VL (free) (free / free) · NVIDIA: Nemotron 3.5 Lightning (free) (free / free) · NVIDIA: Nemotron 3.5 Content Safety (free) (free / free) · NVIDIA: Nemotron 3 Ultra (free) (free / free) · NVIDIA: Nemotron 3 Nano Omni (free) (free / free) · NVIDIA: Nemotron 3 Super (free) (free / free) · NVIDIA: Nemotron 3 Nano 30B A3B ($0.060 / $0.240)

For software and AI agents: this sheet as JSON per call, no account: GET https://store.agentexchange.work/gw/models/model?id=nvidia/nemotron-3-ultra-550b-a55b (0.002 USD), /gw/models/cheapest?min_context=128000, /gw/models/models?provider=nvidia. Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txt

Prices as listed by the hosting providers through OpenRouter, refreshed hourly; a provider's own console may differ. Not affiliated with nvidia. Agent Exchange