NVIDIA officially unveiled the most powerful Nemotron 3 Ultra model in the Nemotron 3 series on June 4, 2026. This massive MoE (Mixture of Experts) model features 550 billion total parameters and 55 billion active parameters, utilizing an innovative Mamba-Transformer hybrid architecture and supporting an ultra-long context of up to 1 million tokens. Now, Nous Research, the developer of Hermes Agent, has announced its joining of NVIDIA’s Nemotron Coalition alliance. In partnership with NVIDIA and the cloud computing platform Nebius, they are offering two weeks of free access to Nemotron 3 Ultra on Nous Portal from June 4 to June 18.
NVIDIA’s Most Powerful Open-Source Model Nemotron 3 Ultra Hermes: Free for Two Weeks Only
This is great news for Hermes Agent users. Previously, we introducedThe Xiaomi MiMo V2 Pro model is available for free for a limited time via Nous PortalThe setup method is the same as before for this operation process, except the free model has been replaced with NVIDIA’s flagship Nemotron 3 Ultra. If you’ve already followed the previous tutorial to set up Nous Portal, you can simply switch the model and start using it directly.
We are excited to join Nvidia’s Nemotron Coalition of leading AI labs working together to advance open frontier foundation models.
To celebrate we have partnered with @nvidia and @nebiustf to provide 2 free weeks of the new Nemotron 3 Ultra model on the Nous Portal! pic.twitter.com/z4jFzPgY17
— Nous Research (@NousResearch) June 4, 2026
Nemotron 3 Ultra: Built for Agent Scenarios
Nemotron 3 Ultra is NVIDIA Nemotron 3 familythe top-tier model among them, positioned as an “Open Frontier Model”. Its core features include:
550B total parameters, 55B active parametersAdopts the MoE (Mixture of Experts) architecture, activating only 55 billion active parameters out of 550 billion during each inference, significantly reducing computational costs.
Mamba-Transformer Hybrid ArchitectureCombining Mamba’s linear attention with Transformer’s precise reasoning capabilities to achieve more efficient inference throughput
5x inference speedCompared to GLM-5.1-754B-A40B and other open-source models of the same tier, Nemotron 3 Ultra delivers 5.9x higher inference throughput.
1M context tokens: Supports ultra-long context processing, surpassing same-tier open-source models on the RULER benchmark
Optimized for Agent scenariosIt excels particularly in agentic tasks such as code writing, long-term planning, and tool calling.
NVIDIA stated on X that Nemotron 3 Ultra’s hybrid architecture can complete more inference cycles within the same time budget, enabling Agents to complete more tasks in less time. For users running complex tasks with the Hermes Agent, this means faster response times and lower usage costs.
Today we’re shipping Nemotron 3 Ultra.
A 550B MoE frontier-intelligence open model built for long-running agents.
It delivers 5x faster inference and lowers the cost of complex agentic tasks by up to 30% versus other open frontier models. pic.twitter.com/FEXqvfzQFO
— NVIDIA AI (@NVIDIAAI) June 4, 2026
Before You Begin
Before you begin, please confirm the following:
Installed Hermes Agent (either desktop or CLI version), have a Nous Portal account (free registration), this free planNo credit card requiredYou can directly choose the Free plan. If you haven’t installed Hermes Agent yet, below we’ll cover how to install both the desktop and CLI versions. If you already have Hermes Agent, you can skip ahead to the “Choose a free model” step.
Method 1: Desktop Version (Simplest)
The desktop version is the simplest way to install, requiring no terminal commands at all. If you’ve already set up the Hermes desktop version before, the second half of this article has the tutorial.
Step 1: Download and Install
Go to the official Hermes Agent download page and download the installer for your operating system (macOS or Windows). After launching it, the program will automatically complete the initial setup, which usually takes less than a minute. For those who are not sure how to do it, you canLook at this one。
Step 2: Connect to Nous Portal
When you first open it, you’ll see the “Let’s get you set up” screen. Tap Nous Portal (marked as Recommended).
Next, the browser will automatically open the Nous Portal page. After creating an account or logging in, select the Free plan and authorize Hermes to connect. The application will automatically complete the binding.
Next, you’ll set up a login account—just choose whichever account you commonly use. Here we’ll select Google, and that completes the connection setup.
Step 3: Choose a Free Model
After the connection is complete, click Change on the Default model card:
Scroll down to search for “nemotron 3 ultra” and select the version marked as Free tiernvidia/nemotron-3-ultra:free⚠️ Note: Be sure to select the one with :free The version with the suffix is the key to the free plan. Selecting other versions will incur charges (though you won’t be able to use them anyway if you haven’t set up a credit card). Nous Portal official frequently has free models available, and if you don’t want to spend money, you can pair them with Openrouter’s free models to save quite a bit.
Step 4: Getting Started
Click “Start chatting” to start using Nemotron 3 Ultra for free.
Method 2: CLI Command Line (Terminal Users)
Users who are comfortable using the terminal can complete the setup by following the steps below.
Step 1: Install Hermes Agent
Enter the following command in the terminal:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
After installation is complete, reload the shell configuration:
source ~/.bashrc(or source ~/.zshrc）
Step 2: Run Quick Setup
hermes setup
Select Quick Setup, and Hermes will automatically open a browser page.
Step 3: Create a Nous Portal Account
Create a Nous Portal account in your browser (or log in to an existing account), and select the Free plan.
Step 4: Connect Account
The browser will prompt you to connect your Nous Portal account to Hermes Agent. After clicking Connect, a confirmation screen will appear.
Step 5: Select Free Model
Go back to the terminal and select from the model list:
nvidia/nemotron-3-ultra:free
Also note that one must choose those with :free Only the version with the suffix can enjoy the free plan.
Step 6: Get Started
After completing the remaining setup steps, enter hermes You can start now and begin chatting with Nemotron 3 Ultra.
How do existing Hermes Agent users switch?
If you’ve already set up Hermes Agent before (for example, if you’ve used MiMo V2 Pro free tier), switching is very simple:
Desktop versionOpen the model selector (Providers), first select “Nous Portal” and search for “nemotron 3 ultra”, then choose the Free tier version, same as above.
– CLI / TUIIn the conversation, tell Hermes to go to the official website to set up the latest version. nemotron-3-ultra:freeInput after the model reports completion /model nvidia/nemotron-3-ultra:free Switch directly or enter /model Open selector to select manually
You can enter after switching is complete /model Verify that the currently used model is correct.
Limited-time free offer—act fast!
Nemotron 3 Ultra’s free period ends on June 18. During this time, you can use NVIDIA’s most powerful open-source model at no cost. For Hermes Agent users, this is a great opportunity to experience a top-tier AI model, and the setup process is straightforward—only takes a few minutes. Users who have already set up Nous Portal will find it even easier, just switch to the model directly.
The Nemotron Coalition has brought together top AI labs including NVIDIA, Nous Research, and Nebius, with the goal of jointly advancing the development of open frontier foundation models. This limited-time free offer for Nemotron 3 Ultra is the first wave of results from this coalition partnership. For everyday users, opportunities to access a 550B parameter-class model without spending a single cent are indeed rare.
Just a reminder to switch back to your original model or set up a Fallback before the free period ends, so you don’t get suddenly disconnected. If you haven’t tried Hermes Agent yet, this Nemotron 3 Ultra limited free offer is also a great opportunity to get started.
Source: KOCPC Chinese