NVIDIA: Nemotron 3 Ultra (free)

nvidia/nemotron-3-ultra-550b-a55b:free

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...

Modalities

In / out price

Free / Free per 1M

Context

Released

Jun 4, 2026

Why use Nemotron 3 Ultra (free)

Step-by-step reasoning mode
Tool and function calling (agent-ready)
Very long context: whole books or codebases in one prompt

Pricing

Input	Free / 1M tokens
Output	Free / 1M tokens

List prices via OpenRouter, checked daily. Real cost depends on the provider and on reasoning tokens.

How it ranks

Intelligence index22.9

Coding index49.3

Agentic index20.1

Source: Artificial Analysis (artificialanalysis.ai) via OpenRouter (openrouter.ai/rankings).

Cost calculator

Pick a task, set how often you run it, and compare what it costs on each model.

TaskInput tokens≈ 602 wordsOutput tokens≈ 1,504 wordsRuns per month

Input served from cache: 0%Repeated prompts (same instructions or document) are billed at the cheaper cache-read price where the model supports it.

Model	Per run	Per month	Per 1,000 runs
NVIDIA: Nemotron 3 Ultra (free)	$0	$0	$0

Estimates from list prices via OpenRouter; real bills vary by provider and reasoning tokens.

Will it fit? Context window simulator

Each square is about one page (500 words). Colored squares are your content; grey squares are free space in the model's context window.

Your content

Novel (90,000 words)

Total: about 120,000 tokens (≈ 90,226 words). Token counts are estimates.

Models

Zoom (square size): 6pxZoom out to see a whole million-token window at once.