NVIDIA Nemotron 3 Super is a 120B-parameter open hybrid MoE model, activating just 12B parameters for maximum compute efficiency and accuracy in complex multi-agent applications. Built on a hybrid Mamba-Transformer...
- Modalities
- text → text
- In / out per 1M
- Free / Free
- Context
- 1M tokens
- Added
- Mar 11, 2026
- Max output
- 262K tokens
- Tokenizer
- Other
- Reasoning
- optional
Pricing
Per 1M tokens. The provider price and our flat 3% fee are separate columns — what you pay is their sum.
| Per 1M tokens | Provider | + 3% fee | You pay |
|---|---|---|---|
| Input | Free | — | Free |
| Output | Free | — | Free |
Providers
One company serves Nemotron 3 Super (free). Figures are each provider’s own list price per 1M tokens — your OpenKey price is in the card above.
| Provider | Context | Input /M | Output /M |
|---|---|---|---|
| Nvidia | 262K | Free | Free |
Provider list from the OpenRouter catalog, Jul 28, 2026. All providers
Benchmarks
Artificial Analysis
- Agentic index
- 8.7
- Coding index
- 37.7
- Intelligence index
- 25.4
Supported parameters
- include_reasoning
- max_tokens
- reasoning
- response_format
- seed
- structured_outputs
- temperature
- tool_choice
- tools
- top_p
Call it
OpenAI-compatible: point your SDK at api.openkey.ai/v1 and use
model nvidia/nemotron-3-super-120b-a12b:free.
curl https://api.openkey.ai/v1/chat/completions \
-H "Authorization: Bearer ***" \
-H "Content-Type: application/json" \
-d '{
"model": "nvidia/nemotron-3-super-120b-a12b:free",
"messages": [{"role": "user", "content": "Hello"}]
}'import os
from openai import OpenAI
client = OpenAI(
base_url="https://api.openkey.ai/v1",
api_key=os.environ["OPENKEY_API_KEY"],
)
completion = client.chat.completions.create(
model="nvidia/nemotron-3-super-120b-a12b:free",
messages=[{"role": "user", "content": "Hello"}],
)
print(completion.choices[0].message.content)import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://api.openkey.ai/v1",
apiKey: process.env.OPENKEY_API_KEY,
});
const completion = await client.chat.completions.create({
model: "nvidia/nemotron-3-super-120b-a12b:free",
messages: [{ role: "user", content: "Hello" }],
});
console.log(completion.choices[0].message.content);About Nemotron 3 Super (free)
Nemotron 3 Super (free) is a 120B-parameter hybrid Mamba-Transformer MoE model that activates only 12B parameters per token, which is what keeps compute cost down while targeting complex multi-agent workloads. It ships a 1M-token context window and caps completions at 262,144 tokens. On Artificial Analysis benchmarks it scores 25.4 on the intelligence index, 37.7 on coding, and 8.7 on agentic tasks — text-in, text-out only, no vision.
This is a free variant of NVIDIA's paid nemotron-3-super-120b-a12b (which lists at $0.085/M input, $0.4/M output), so you get the same 12B-active-parameter architecture at zero cost. Its 1M-token context window beats 78% of catalog models. It supports optional reasoning (medium or low effort, on by default) plus tool calling and structured outputs — a combination most free models in the catalog don't offer.
Questions
- How much does Nemotron 3 Super (free) cost via API?
- It's free — $0 per 1M input tokens and $0 per 1M output tokens, with no OpenKey fee applied since 3% of zero is zero. The paid version of this same architecture, nvidia/nemotron-3-super-120b-a12b, charges $0.085/M input and $0.4/M output before OpenKey's fee.
- What is Nemotron 3 Super (free)'s context window?
- It supports a 1,000,000-token context window, larger than 78% of models in the catalog. That's enough to hold roughly 750,000 words of text in a single request, well past what's needed for most document or codebase analysis tasks. Max completion length is capped at 262,144 tokens.
- Is Nemotron 3 Super (free) actually free to use?
- Yes — both prompt and completion pricing are $0.0 per 1M tokens, and it's one of 25 free models available on OpenKey. Unlike the paid nvidia/nemotron-3-super-120b-a12b variant, this version carries no per-token cost while running the same 12B-active-parameter MoE architecture.
- Does Nemotron 3 Super (free) support tool calling and structured outputs?
- Yes, it supports tools, tool_choice, response_format, and structured_outputs, alongside reasoning, seed, and temperature parameters. Reasoning is optional but enabled by default at medium effort, with low effort also available. There's no vision support — input and output are text only.
- How does Nemotron 3 Super (free) compare to Llama 3.3 Nemotron Super 49B V1.5?
- Llama 3.3 Nemotron Super 49B V1.5 costs $0.4/M for both input and output tokens, while Nemotron 3 Super (free) costs nothing. The two also differ architecturally: the 49B model activates its full parameter set per token, while this 120B MoE model activates only 12B parameters per token, trading a larger total model for lower compute per request.
More from NVIDIA
All NVIDIA models →
NVIDIANemotron 3 Ultra
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...
ctx 1M in $0.515 out $2.27
NVIDIAFreeNemotron 3 Ultra (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...
ctx 1M in Free out Free
NVIDIAFreeNemotron 3.5 Content Safety (free)
NVIDIA Nemotron 3.5 Content Safety is a compact 4B-parameter multimodal guardrail model from NVIDIA, fine-tuned from Google Gemma-3-4B. It moderates both inputs to and responses from LLMs and VLMs, accepting...
ctx 128K in Free out Free
NVIDIAFreeNemotron 3 Nano Omni (free)
NVIDIA Nemotron™ 3 Nano Omni is a 30B-A3B open multimodal model designed to function as a perception and context sub-agent in enterprise agent systems. It accepts text, image, video, and...
ctx 256K in Free out Free