Nemotron 3 Super is Live on FriendliAI: multi-agent applications and for specialized agentic AI systems

FriendliAI is launching Day-0 support for NVIDIA Nemotron 3 Super. Our Dedicated Endpoints are optimized to unlock the full efficiency of Nemotron’s hybrid MoE architecture—delivering high-throughput inference, low-latency execution, and enterprise-ready deployment for multi-agent and tool-using AI systems. From evaluation to production, FriendliAI provides the performance infrastructure required to scale Nemotron 3 Super in real-world agentic workloads.

NVIDIA Nemotron 3 Super is NVIDIA’s next-generation open reasoning model, purpose-built for multi-agent systems and specialized agentic AI. Powered by a hybrid Transformer–Mamba MoE architecture, it combines high accuracy with exceptional compute efficiency, unlocking tool-using agents and long-context workflows at production scale.

Provision your endpoint here.

NVIDIA Nemotron 3 Super Provides:

What can you build with Nemotron 3 Super on FriendliAI?

Nemotron 3 Super is specifically engineered for agentic intelligence, multi-agent coordination, tool interactions, and long-context reasoning, and FriendliAI’s high-performance hosting unlocks these capabilities at production scale. With Nemotron 3 Super on FriendliAI, developers can build applications that go beyond static generation and into dynamic, actionable AI workflows.

Here are a few powerful categories of next-generation applications you can build:

Friendli Dedicated Endpoints for NVIDIA Nemotron 3 Super

FriendliAI enables production deployment of NVIDIA Nemotron 3 Super with dedicated GPUs, predictable performance, and 99.99% uptime SLA, so teams can scale agentic workloads without managing infrastructure.

Built for enterprise AI systems, Friendli Dedicated Endpoints support multi-agent workflows, tool calling, and long-context reasoning with optimized MoE inference and reserved GPU capacity.

Key Capabilities

• Private Nemotron 3 Super deployment on dedicated GPUs for consistent performance
• High-throughput, low-latency MoE-optimized inference to maximize token throughput
• Production-ready endpoints with built-in observability and autoscaling
• Flexible GPU options (B200, H100/H200) based on performance needs
• Enterprise-grade reliability, security, and full data ownership in a private environment

FriendliAI handles serving, batching, scaling, and GPU orchestration, letting your team focus on building, not infrastructure.

With Friendli Dedicated Endpoints, teams can:

• Deploy Nemotron 3 Super into production within minutes
• Scale multi-agent systems without rebuilding infrastructure
• Run long-context reasoning pipelines with predictable latency
• Effortless autoscaling: scale inference dynamically across GPUs, instantly right-sizing capacity to match demand.
• Support tool-calling workflows at enterprise scale
• Operate customized Nemotron models securely and privately

FriendliAI removes the operational complexity of serving frontier models — enabling developers to focus on agentic intelligence, not infrastructure.

Get started with Nemotron 3 Super on FriendliAI

To deploy NVIDIA Nemotron 3 Super on Friendli Dedicated Endpoints,

1️⃣ Navigate to the dedicated endpoint creation page.

2️⃣ Choose your desired model, such as "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16".

3️⃣ Click “Create”

You can send requests to Nemotron 3 Super using any OpenAI-compatible inference API/SDK. For example, using Friendli Python SDK:

Prerequisites

Before getting started, you need to set up:

• A FriendliAI account
• Friendli Token from Friendli Suite settings tab

Install the package

uv pip install friendli

Environment Setup

Set up your FriendliAI API key (aka Friendli Token):

export FRIENDLI_TOKEN="your-token-here"

Example Code

import os

from friendli import SyncFriendli

with SyncFriendli(
    token=os.environ["FRIENDLI_TOKEN"],
) as friendli:
    res = friendli.dedicated.chat.stream(
        model="your-dedicated-endpoint-id",
        messages=[
            {"role": "system", "content": "You are a helpful assistant."},
            {"role": "user", "content": "Hello, Nemotron 3 Super!"},
        ],
    )
    for chunk in res:
        if content := chunk.data.choices[0].delta.content:
            print(content, end="")

Run better with FriendliAI

Nemotron 3 Super is redefining what’s possible with open reasoning models and FriendliAI is your platform to take it to production. From experimentation to enterprise deployment, FriendliAI delivers the performance, reliability, and control required to scale agentic AI in real-world applications.

👉 Launch your Nemotron 3 Super Dedicated Endpoint today and start building next-generation agentic AI with FriendliAI.