Skip to content
RankShark

Nemotron 3 Nano 30B A3B

NVIDIA Nemotron 3 Nano 30B A3B is a small language MoE model with highest compute efficiency and accuracy for developers to build specialize

ModelOpen sourceby NVIDIAsince 2025
47.9

SharkScore

#70 in Models

Why it scores 47.9

Nemotron 3 Nano 30B A3B has a SharkScore of 47.9/100, ranked #70 of 608 in Models (updated 2026-11-19).

Each part compares Nemotron 3 Nano 30B A3B with other listings in Models. Evidence covers 87% of the score's weight.

Adoption
35 · 30%
User reviews
no data · 25%
Maintenance
42 · 20%
Trust & transparency
80 · 15%
Momentum
no data · 10%

Paying never changes this score. How ranking works · method v2026-11-18.2

RankShark review of Nemotron 3 Nano 30B A3B

Automated evaluation by RankShark · 18 Nov 2026 · method v1.0 · How we review

We were not paid for this review.

As of 2026-11-18, Nemotron 3 Nano 30B A3B is a text-only reasoning and chat model; OpenRouter lists a 262,144-token context at $0.06/$0.24 per million input/output tokens. [E1, E2, MEASURED]

NVIDIA’s 30B-parameter text model uses a hybrid Mamba-2 and mixture-of-experts architecture, with 3.5B active parameters. It is designed for reasoning, chat, coding, and agent applications, and supports six languages. [E1]

Best for

  • Developers building text-based AI agents, chatbots, or RAG systems. · model card
  • Developers who need reasoning and non-reasoning modes, with an option to omit intermediate reasoning traces. · model card
  • Teams working in English, German, Spanish, French, Italian, or Japanese. · model card

Not for

  • Applications that require image, audio, or video input or output; the listed model accepts and returns text. · model card, OpenRouter

Key features

  • Hybrid model architecture. The vendor says the model combines Mamba-2, attention, and mixture-of-experts layers, with 3.5B active parameters out of 30B total. · model card
  • Configurable reasoning. The vendor says reasoning traces can be enabled or omitted through a chat-template flag; omitting them may reduce accuracy on harder reasoning prompts. · model card
  • Language support. The vendor says the model supports English, German, Spanish, French, Italian, and Japanese. · model card
  • Agent and tool-use training. The vendor says post-training included tool calling and reinforcement learning across multi-step tool-use and other task environments. · model card
  • OpenRouter context and token rates. RankShark measured a 262,144-token context and rates of $0.06 per million input tokens and $0.24 per million output tokens on the OpenRouter listing. · OpenRouter

Pros

  • The model supports six named languages and text input and output. · model card, OpenRouter
  • The vendor describes a configurable reasoning mode and training for tool calling and multi-step tool use. · model card
  • The vendor provides model weights and describes the model as ready for commercial use under the NVIDIA Nemotron Open Model License. · model card

Cons

  • The listed model is text-only, so it does not fit applications requiring non-text modalities. · model card, OpenRouter
  • The Hugging Face page states a maximum input size of 1M tokens, while the OpenRouter listing and RankShark measurement show a 262,144-token context; check the limit for the deployment route you use. · model card, OpenRouter
  • No free plan or free trial stated on the pages we checked.
  • Whether customer data is used for model training is not stated on the pages we checked.

Pricing

PlanPriceLimits
OpenRouter listing (usage-based)$0.06 per million input tokens; $0.24 per million output tokens262,144-token context window, measured on 2026-11-18

Checked 18 Nov 2026

Free plan
Not stated (checked 18 Nov 2026)
Free trial
Not stated (checked 18 Nov 2026)
Trains on customer data
Not stated (checked 18 Nov 2026)
Data retention
Not stated (checked 18 Nov 2026)
Data regions
Not stated (checked 18 Nov 2026)
Data processing agreement
Not stated (checked 18 Nov 2026)
AI models used
NVIDIA Nemotron 3 Nano 30B A3B; 30B total parameters and 3.5B active parameters. · model card, OpenRouter
API
The OpenRouter listing shows chat completions, completions, messages, and responses endpoints. · OpenRouter
SDKs
The model is integrated into Hugging Face Transformers from v5.3.0; the vendor lists NeMo 25.11.01 as a runtime engine. · model card
Webhooks
Not stated (checked 18 Nov 2026)
Platforms
The vendor lists Linux and NVIDIA H100-80GB and A100 hardware compatibility. The model is listed on Hugging Face and OpenRouter. · model card, OpenRouter
Security claims
No security certification stated (checked 18 Nov 2026)

What we measured

Context window (OpenRouter listing) · now
262,144 tokens
Price per million tokens (OpenRouter listing) · now
$0.06 in, $0.24 out
Package downloads · 30 days
1,334,806
How we tested · sources
  • Hourly uptime and response-time checks of the product's URL.
  • Repository, release and download data from GitHub, npm, PyPI and Hugging Face where they exist.
  • We did not sign up, pay, or use the product as a customer.
  • 18 Nov 2026: Plans or prices changed on the pricing page.
  • 18 Nov 2026: Plans or prices changed on the pricing page.

Something wrong? Report an error with a public link that shows it; corrections re-run the review.

Top alternatives to Nemotron 3 Nano 30B A3B

See all
  • 67.1

    Qwen3.8 27B is an open-weight dense vision-language model from Qwen.

    ModelOpen source
  • 65.5

    Qwen3.6-35B-A3B is an open-weight multimodal model from Alibaba Cloud with 35 billion total parameters and 3 billion active parameters per t

    ModelOpen source
  • 64.7

    Qwen3.8 Flash is a multimodal reasoning model from Alibaba.

    ModelOpen source
  • 64.3

    Qwen3.6 27B is a dense 27-billion-parameter language model from the Qwen Team at Alibaba, released in April 2026.

    ModelOpen source