Nemotron 3 Super (free)
Nemotron 3 Super (free) is NVIDIA’s open‑weights, high‑throughput 120B-parameter hybrid mixture‑of‑experts language model, optimized for complex agentic AI and multi‑agent reasoning workloads.
What is Nemotron 3 Super (free)?
Nemotron 3 Super is an open, 120B-parameter hybrid Mamba-Transformer mixture-of-experts model from NVIDIA designed for high-accuracy, efficient agentic reasoning. It is mainly used to power multi-agent and enterprise AI workflows that require long-context reasoning, planning, and orchestration across many tools or services. It is also well-suited for code, math, and complex multistep generation tasks where high throughput and long sequences are important. It belongs to the Nemotron 3 family of open models (Nano, Super, Ultra), succeeding earlier Nemotron generations.
Providers
Different companies host the same model. OpenRouter routes your request to one of them based on the routing mode you pick — Balanced (price + speed), Nitro (fastest), or Exacto (highest tool-calling accuracy).
| Provider | Input | Output | Cache read /M | Latency | Throughput | Uptime |
|---|---|---|---|---|---|---|
| NVIDIA | $0.00 | $0.00 | — | ~180ms | ~60 tps | ~99.9% |
| AWS Bedrock | ~$0.25 | ~$0.25 | — | ~220ms | ~45 tps | 100.00% |
| Google Cloud | ~$0.24 | ~$0.24 | — | ~210ms | ~50 tps | ~99.9% |
| Azure AI | ~$0.26 | ~$0.26 | — | ~230ms | ~40 tps | 100.00% |
Try this model
Test Nemotron 3 Super (free) right here — free to start.
Suggestions for your first prompt
Code snippet
Call the model through the OpenAI-compatible API.
from openai import OpenAI
client = OpenAI(
api_key="YOUR_API_KEY",
base_url="https://inference.example.com/v1"
)
response = client.chat.completions.create(
model="nvidia/nemotron-3-super-free",
messages=[
{
"role": "user",
"content": "Describe this image in one sentence."
}
],
)
print(response.to_json())
{
"model": "nvidia/nemotron-3-super-free",
"messages": [
{
"role": "user",
"content": "Describe this image in one sentence."
}
]
}
5 Core Capabilities
-
Agentic Reasoning
Supports multi‑agent, tool-using AI workflows, coordinating complex tasks with high throughput and long-horizon reasoning across agents.
-
Long-Context Processing
Handles sequences up to around one million tokens, enabling analysis of large documents, codebases, and extended conversations without losing context.
-
Multilingual Text
Generates and understands text in multiple languages, including English and Japanese, for global applications and cross-lingual workflows.
-
General Chat
Engages in open-domain dialogue, following instructions, answering questions, and assisting with writing or brainstorming in natural language.
-
Code and Data Text
Trained on diverse web, code, and technical data, enabling structured outputs, explanations, and reasoning over text-based information sources.
6 Most Valuable Use Cases
- Code Generation Assistance
- Customer Support Chatbots
- Document Summarization
- Semantic Text Tagging
- Legal Case Research
- Regulatory Case Monitoring
Why Build on LLM.API?
One unified API. Every major model. Built-in reliability, cost control, and observability.
-
Unified AI Routing
Intelligently route each request across providers and models based on latency, capability, or custom rules. One API, always the best path for your workload.
Smart traffic, single endpoint -
Cost-Aware Orchestration
Automatically balance performance and price with configurable policies. Use premium models when it matters, fall back to cheaper ones when it doesn’t.
Optimize spend by default -
Resilient Fallbacks
Define multi-step failover chains across providers so requests keep flowing through outages, rate limits, or model errors—without touching your application code.
Stay online under stress -
Deep Observability
Get full visibility into requests, tokens, latency, errors, and providers with structured logs and traces. Debug faster and tune workloads with real data.
See every token spent -
Task-Level Abstractions
Describe tasks like chat, tools, reranking, or extraction once and run them on any model. Ship features without rewriting prompts per provider.
Code to tasks, not models -
High-Throughput Batch
Submit massive batch jobs through a single API with queuing, retries, and cost controls built-in. Process millions of inputs without custom infrastructure.
Scale jobs, not ops
When to Use — When NOT to Use
Use it if...
- You need a free, general-purpose model for everyday coding, writing, and Q&A.
- You need to prototype AI features without incurring usage costs during experimentation.
- Your use case involves moderate-length chats where perfect reasoning is not critical.
- Your use case involves simple code snippets, bug fixes, or small refactors.
- You need a baseline model to compare against stronger proprietary or paid systems.
- Your use case involves occasional content generation, summaries, and simple data extraction.
Avoid if...
- You need state-of-the-art reasoning quality for complex multi-step or high-stakes decisions.
- Your workload requires very long-context processing, such as full-book analysis or logs.
- You need top-tier code generation for large projects, architectures, or unfamiliar stacks.
- Your workload requires highly reliable factual answers on niche, technical, or evolving topics.
- You need best-in-class safety controls, compliance certifications, or robust content-filter customization.
- Your workload requires highly optimized latency and throughput for large-scale, performance-critical production.
Frequently Asked Questions
-
What is Nemotron 3 Super (free)?
Nemotron 3 Super (free) is an NVIDIA large language model accessible via LLM.API, tuned for general-purpose text generation and assistant-style conversations.
-
What is Nemotron 3 Super (free) best suited for?
It is best for fast, low-cost chat-style interactions, drafting content, and lightweight reasoning where cost and accessibility matter more than cutting-edge intelligence.
-
How is Nemotron 3 Super (free) priced on LLM.API?
The free tier incurs no direct per-token charges to you, but may be subject to rate limits and usage caps enforced by LLM.API.
-
What context window does Nemotron 3 Super (free) support?
Nemotron 3 Super (free) supports a context window of up to 8K tokens, including both prompt and response tokens.
-
How fast is Nemotron 3 Super (free) in terms of latency?
Latency is typically low for short prompts, but can increase under heavy shared-load conditions because the free tier runs on pooled infrastructure.
-
Which modalities does Nemotron 3 Super (free) support?
Nemotron 3 Super (free) supports text-in, text-out interactions only and does not natively process images, audio, or video.
-
How do I call Nemotron 3 Super (free) through the LLM.API gateway?
You select the model by its identifier in the LLM.API completion or chat endpoint, passing your prompt and standard configuration parameters like temperature.
-
How does Nemotron 3 Super (free) compare to larger NVIDIA or frontier models?
Compared to larger or paid frontier models, it is generally cheaper and more accessible but weaker on complex reasoning, coding, and long-context tasks.
-
What are the main limitations of Nemotron 3 Super (free)?
It may hallucinate facts, struggle with very long or deeply technical tasks, and lacks multimodal capabilities and fine-grained enterprise controls.
-
Can I use Nemotron 3 Super (free) for production workloads?
You can, but should account for potential rate limits, variable performance, and weaker reliability than dedicated, paid production-grade NVIDIA deployments.
COMPARE
Competitive Models
-
Claude Opus 4.6
Claude Opus 4.6 is a large language model from Anthropic’s Claude Opus series, designed as a high-end, general-purpose AI assistant with strong reasoning and language capabilities. It is notable for being one of Anthropic’s flagship frontier models, aimed at complex tasks requiring advanced comprehension and generation.
-
GLM 4.6V
GLM 4.6V is Z.ai’s open-source, large-scale vision-language model that supports images, video, documents, and text with a long context window and native tool use. It is notable for combining high-quality multimodal understanding with function calling and cloud- or local-friendly variants.
-
GPT-4o Transcribe
GPT-4o Transcribe is an OpenAI model specialized for converting audio into accurate, time-aligned text transcripts. It is notable for handling natural speech, varied accents, and real-world audio conditions with high reliability.
Get one key to every model
Swap your API key. Keep your code.