nemotron-3-super-120b-a12b
NVIDIA Nemotron 3 Super is a 120B-parameter open hybrid MoE model that activates just 12B parameters, built on a hybrid Mamba-Transformer design for complex multi-agent applications.
Pricing: Input 15 credits / 1M tokens (≈ $0.07), Output 80 credits / 1M tokens (≈ $0.40).
Prices updated
How can I help you today?
This demo page does not include the full memory features of the client. For the complete experience, please use the client or the CLI.
More AI models from NVIDIA
Latest and popular AI models
Nemotron 3 Super API on ModelFold
NVIDIA Nemotron 3 Super is a 120B-parameter open hybrid MoE model that activates just 12B parameters, built on a hybrid Mamba-Transformer design for complex multi-agent applications. Try Nemotron 3 Super in the ModelFold Playground, then call it from your own code through an OpenAI-compatible chat completions API.
What Is the Nemotron 3 Super API
Nemotron 3 Super at a glance
Nemotron 3 Super is a chat model from NVIDIA. It accepts text as input and returns text, with a context window of up to 262K tokens.
Use it without hosting anything
On ModelFold you call Nemotron 3 Super with a single API key — no model weights to download and no inference servers to run. The same key also works for the other chat, image, video and audio models on ModelFold.
Core Capabilities of Nemotron 3 Super
Up to 262K tokens of context
Keep long conversations, large documents and sizeable code in a single request, so the model can reason over the full material instead of fragments.
Function calling
Nemotron 3 Super supports tool (function) calling: describe your functions in the request and the model returns structured calls your code can run — the basis for agents and integrations.
Adjustable reasoning effort
Choose a reasoning effort level (low, medium or high) per request to balance depth of thinking against speed and token usage.
Structured outputs
Ask for JSON that follows a schema (response_format), which makes extraction, classification and system integrations more reliable.
How to Use Nemotron 3 Super on ModelFold
- 1
Step 1: Try it in the Playground
Open the Playground at the top of this page, type a prompt and see how Nemotron 3 Super responds before writing any code.
- 2
Step 2: Create an API key
Sign in to ModelFold and create an API key in the console. Keep it secret and send it as a Bearer token.
- 3
Step 3: Send a request
Call POST https://modelfold.ai/v1/chat/completions with "model": "nemotron-3-super-120b-a12b" and your messages. Existing OpenAI-compatible SDKs work by changing the base URL and key.
- ✓
Step 4: Ship and monitor
Integrate the response into your product and check token usage and spend for every call in the console logs.
What You Can Build with Nemotron 3 Super
Autonomous agents
Build agents that plan multi-step tasks, call your tools and APIs, and keep working toward a goal across long sessions.
High-volume pipelines
Run classification, extraction, rewriting and batch jobs at scale where cost per request matters.
Analysis and complex reasoning
Work through research questions, data interpretation, planning and other problems that need careful step-by-step reasoning.
Tool-calling backends
Route requests, call functions and chain API calls as the worker model inside larger agent systems.
Frequently Asked Questions
What is Nemotron 3 Super?⌄
NVIDIA Nemotron 3 Super is a 120B-parameter open hybrid MoE model that activates just 12B parameters, built on a hybrid Mamba-Transformer design for complex multi-agent applications.
Which inputs does Nemotron 3 Super support?⌄
Nemotron 3 Super accepts text and returns text. Through ModelFold you send text messages; documents attached in the Playground are converted to text.
How long is the context window of Nemotron 3 Super?⌄
Up to 262K tokens, shared between your input and the model’s output.
Does Nemotron 3 Super support function calling?⌄
Yes. Nemotron 3 Super supports tools / function calling through the standard tools parameter of the chat completions API.
How is Nemotron 3 Super billed on ModelFold?⌄
Pay as you go by tokens: input and output tokens are charged at the per-million-token prices shown on this page. There is no monthly fee, and failed requests are not charged.
How do I call Nemotron 3 Super through the API?⌄
Send a POST request to https://modelfold.ai/v1/chat/completions with "model": "nemotron-3-super-120b-a12b", using your ModelFold API key as the Bearer token.
| Input | 15 credits 1M tokens≈ $0.07 |
| Output | 80 credits 1M tokens≈ $0.40 |
Billed per token: input and output are priced separately, and thinking tokens count as output. The Playground charges exactly the same way.
Read API docs →AI Agents · Claude Code / Cursor / Trae →curl https://modelfold.ai/v1/chat/completions \
-H "Authorization: Bearer $MODELFOLD_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"nemotron-3-super-120b-a12b","messages":[{"role":"user","content":"Hello"}]}'