Nemotron 3 Nano Omni 30B A3B Reasoning
Neural Network
Nemotron 3 Nano Omni 30B A3B Reasoning is an open NVIDIA multimodal model with a reasoning mode for agentic systems.
Specifications
- Input
- text, images, documents
- Output
- text
- Context size
- 256 000 tokens
- Max answer length
- 65 536 tokens
- Reasoning effort
- low, medium, high
- Capabilities
- reasoning, function calling
How it works Nemotron 3 Nano Omni 30B A3B Reasoning?
Nemotron 3 Nano Omni is an open NVIDIA multimodal model with 30 billion parameters, using about three per step. It accepts text, images, and video, designed as a perceptive sub-agent that gathers and prepares context for corporate agentic systems. The reasoning variant works in reasoning mode, breaking down tasks step-by-step before formulating an answer. In BotHub, the model is accessible without a VPN or foreign card; pay with a Russian card in rubles only for actual usage, not by subscription. Other models remain available in the same window for easy comparison. A unified OpenAI-compatible API allows connecting Nemotron to your services and switching to another model later without rewriting the integration, with chats encrypted via AES-GCM. Useful for analyzing document or screenshot content, describing video fragments, gathering context for higher-level agents, running analytical tasks with step-by-step breakdowns, or testing an open model as part of a multimodal pipeline.Frequently asked questions about Nemotron 3 Nano Omni 30B A3B Reasoning
You can use generated results for commercial purposes. You own all rights to the content you create. The only restriction: make sure your prompt does not include copyrighted third-party material. You are responsible for respecting the rights to any input data.
This is an NVIDIA text model with a reasoning mode. It analyzes documents and images, writes and edits code, explains errors, helps with STEM tasks, and calls external functions. A 256,000-token context allows you to work on a large project or a long report in its entirety.
Up to 65,536 tokens in a single response — that's dozens of pages of text or a large code module. If the task is larger, break it into parts: the 256,000-token context allows you to pass everything previously written to the model and continue without losing coherence.
Yes, reasoning before answering is explicitly stated — it is reflected in the name. The model first builds a chain of steps and then provides the result, so it handles math, logic, code analysis, and multi-step questions better, where the breakdown is more important than a guess.
Yes, it supports both function calling and structured JSON output. The model is available via the unified OpenAI-compatible BotHub API, making it easy to integrate into agents and pipelines, and to replace with another model if needed without rewriting the integration.
Images and documents — yes: you can send a screenshot, diagram, photo, or file and ask to analyze the content. Audio and video input is not noted in our data, so for such tasks, it is safer to choose a specialized model in the same window.
We have no data on language support for this model, so we won't make any promises — test it with your task. In BotHub, it's quick: interface and payment in rubles, access without VPN or foreign cards, and dozens of alternatives nearby in the same window.