Yoshi Suhara

Building Small Language Models @nvidia

[Oleksii Kuchaiev @kuchaev]

Banning open-source AI would be a historic mistake — and a self-inflicted wound to U.S. AI leadership.
I've been in this field long enough to remember when GPT-2 was called "too dangerous to release." That didn't age well — and it drew heavy criticism from researchers even at the time. 1/5

[Yoshi Suhara @suhara]

The model was compressed from Nano 9B v2 using Nemotron Elastic (https://t.co/rGOFHEXhYK) and post-trained on Nemotron 3 post-training data using a new recipe designed to improve accuracy in reasoning-off mode. Please see the blog and the Nemotron Elastic paper for details.

[Yoshi Suhara @suhara]

Nemotron 3 Nano 4B is here! A compact hybrid Mamba-Transformer model optimized for edge AI, with top-tier instruction following, strong tool use, and low-VRAM efficiency across NVIDIA Jetson, Spark, and RTX platforms.

HF 🤗 https://t.co/htGZlRPXbq
Blog📝 https://t.co/y3luc63ogU

[Yoshi Suhara @suhara]

Quantized variants are also available!

FP8 🤗 https://t.co/P2W6CNX2hP
GGUF (Q4_K_M) 🤗https://t.co/Kbg40x5nsh

[NVIDIA AI Developer @NVIDIAAIDev]

Introducing NVIDIA Nemotron 3 Super 🎉

Open 120B-parameter (12B active) hybrid Mamba-Transformer MoE model

Native 1M-token context

Built for compute-efficient, high-accuracy multi-agent applications

Plus, fully open weights, datasets and recipes for easy customization and deployment. 🧵

[NVIDIA Japan @NVIDIAJapan]

【新モデル公開🚀Nemotron-Nano-9B-v2-Japanese】

本日、NVIDIA は Nejumi Leaderboard 4 のパラメータ数 10B 以下において、最先端の性能 (SOTA) を達成した NVIDIA Nemotron-Nano-9B-v2-Japanese を公開しました。商用利用可能です。
https://t.co/XEL0h5s1rc

[Pavlo Molchanov @PavloMolchanov]

🚀 New NVIDIA report: NVFP4 + Quantization-Aware Distillation (QAD)

FP4 inference without quality collapse.

Key idea: distill a BF16 teacher into an NVFP4 student using KL loss - much more robust than PTQ/QAT, especially after SFT/RL.

🔥 Near-BF16 accuracy
⚡ ~2-3× throughput, ~1.8× memory savings vs FP8
🧠 Works for LLMs and VLMs (Nemotron Nano, Super, VL)

Technical report: https://t.co/2s72KPCfLO
Research blog: https://t.co/hsBao2KqLw
Hugging Face models: https://t.co/VmZ25qk3Wk

[Jian Zhang @JianZhangCS]

🚀@Nvidia Nemotron 3 Nano is live!

Nemotron 3 Nano is the world's most efficient open MoE with an Hybrid-MoE architecture and 1M context length.

🔥 Strong in reasoning, agentic and chat tasks with leading accuracy among AA index, Tau2, SWE Bench.

🔥 Up to 3.3X higher throughput comparing to other open MoE at similar sizes

🔥 A fully open recipe with data, infra released to the community

Checkout the new model architecture and reinforcement learning technologies we used below:

😊 Huggingface: https://t.co/UX9L9QmuWJ

📢 Research blog: https://t.co/NeTb5xANxR

🛣️Nemo RL & Nemo Gym (RL environment orchestration): https://t.co/fD78eabCZv & https://t.co/E3Q67AIA4j

Kudos to the teams for months of hard work! We are excited to keep building the Nemotron 3 model family and empower the community.

[Georgi Gerganov @ggerganov]

In collaboration with NVIDIA, the new Nemotron 3 Nano model is fully supported in llama.cpp

Nemotron 3 Nano features an efficient hybrid, Mamba, MoE architecture. It's a promising model, suitable for local AI applications on mid-range hardware. The large context window makes it a great choice for a variety of use cases and applications.

The efficiency of llama.cpp and the unique context management features of the `llama-server` tool allows us to deploy and use this model on a wide-range of hardware. With recent code contributions by engineering teams at NVIDIA and open-source collaborators, we can run this model very efficiently across the entire spectrum of NVIDIA GPUs. Learn more at @NVIDIA_AI_PC

https://t.co/3c9LRmfmRp

[Bryan Catanzaro @ctnzr]

Today, @NVIDIA is launching the open Nemotron 3 model family, starting with Nano (30B-3A), which pushes the frontier of accuracy and inference efficiency with a novel hybrid SSM Mixture of Experts architecture. Super and Ultra are coming in the next few months.

[Yoshi Suhara @suhara]

I’m proud to be part of this amazing cross-team, cross-country collaboration supporting Sovereign AI for Japan 🇯🇵!

HF 🤗: https://t.co/436GFaJWPS
Blog 📝: https://t.co/zFKew3HHAo

[NVIDIA Japan @NVIDIAJapan]

🚀 新データセット公開: Nemotron-Personas-Japan

日本の人口構成や文化に合わせて作られた、初のオープンな日本語合成データセットを公開しました。商用利用も可能です。

https://t.co/cVnHcUGvns

[Shizhe Diao @shizhediao]

Thrilled to share that CLIMB has been accepted to the NeurIPS DB track! 🍀 Feeling so lucky to work with such an amazing team. #NeurIPS2025

[Artificial Analysis @ArtificialAnlys]

NVIDIA has released Nemotron Nano 9B V2, a small 9B reasoning model that scores 43 on the Artificial Analysis Intelligence Index, the highest yet for <10B models

Nemotron 9B V2 is the first Nemotron model pre-trained by @NVIDIA. Previous Nemotron models have been developed by post-training on Meta Llama models.

Architecture & Training: The model uses a hybrid Mamba-Transformer architecture. NVIDIA pre-trained a 12B parameter base model and applied post-training with a range of techniques including RLHF and GRPO. The final 9B size was pruned from this model and re-trained with the base model as a teacher.

Small-model frontier: with only 9B parameters, Nemotron Nano 9B V2 is placed ahead of Llama 4 Maverick on our leaderboard, equal to Solar Pro 2 with reasoning and trails just behind gpt-oss-20B (high).

Along with this model, NVIDIA released a 6.6-trillion token subset of their pre-training data for public use on @huggingface

Key model details:
➤ 128k token context window
➤ Supports reasoning and non-reasoning modes (with ‘/no_think’ settings in the system prompt)
➤ Released under the NVIDIA Open Model License, and not additionally covered by Meta’s Llama license like prior Nemotron models - this means that there is no limitation on use by large companies or requirement to keep ‘Nemotron’ in the name of derivative models
➤ No serverless inference providers are yet serving the model, but it is available now on Hugging Face for local inference or self-deployment

See below for our full analysis and key announcement links from NVIDIA 👇

[Yoshi Suhara @suhara]

RT @shizhediao: ✨ Alongside NVIDIA-Nemotron-Nano-v2-9B, we’re also open-sourcing its pre-training dataset.

At NVIDIA, we remain committed…

[NVIDIA AI Developer @NVIDIAAIDev]

We're excited to share leaderboard-topping 🏆 NVIDIA Nemotron Nano 2, a groundbreaking 9B parameter open, multilingual reasoning model that's redefining efficiency in AI and earned the leading spot on the @ArtificialAnlys Intelligence Index leaderboard among open models within the same parameter range.

It's built on a unique hybrid Transformer-Mamba architecture, a combination that delivers the same accuracy you expect, but with higher throughput. This enables it to achieve high performance/cost, making it perfect for real-world applications like customer service agents and chatbots.

🏗️ Hybrid Architecture: By combining the strengths of Transformer and Mamba architectures, achieves up to 6X faster throughput compared to other 8B open models and highest reasoning accuracy.
🏦 Thinking Budget: Reduces unnecessary token generation to cut costs by up to 60%, making it an ideal solution for balancing performance and total cost of ownership (TCO).
🔢 Open Datasets: The training datasets of this model are fully open, giving maximum transparency in using the model for enterprise applications.

🤗 Technical details on @HuggingFace ➡️ https://t.co/Dc87rpUGgj

🏆 Leaderboard ➡️ https://t.co/B9QulXU7iI

[Oleksii Kuchaiev @kuchaev]

We are excited to release Nvidia-Nemotron-Nano-V2 model! This is a 9B hybrid SSM model with open base model and training data. This model also supports runtime "thinking" budget control. HF collection with base and post trained models: https://t.co/n3M01d8lSm

[Pavlo Molchanov @PavloMolchanov]

📢New efficient Hybrid-SLM from NVIDIA-Nemotron-Nano-v2-9B:
❗️6x faster than Qwen3-8B because of Hybrid (Mamba2+Attention) design.

We tried something new: pretrain & align a 12B reasoning model → compress to 9B.

First real stab at reasoning-model compression.

Key takeaways from compression:
▪️Target was 23GB GPUs + room for a 650M vision encoder → design via compression, not bespoke architecture.
▪️Distillation loss went down, but benchmarks didn’t - unlike Base compression.
▪️Reasoning compression needs light post-training alignment.
▪️Applied both Minitron + Puzzle.
▪️Dropped 2 attn layers to hit 128k context; KV cache dominated.
▪️Depth: 62 → 56 (fewer tanked accuracy).
▪️FFN: 20,480 → 15,680 (−23%).
▪️Hidden dimension: 5120 → 4480.
▪️Mamba heads → small gains <15%, mostly avoided.
Distilled on 136B tokens, context grown 8k → 262k. It is required if we want to preserve long-context of 128k.

📰 Report: https://t.co/ypVIcnEa0U
🤗 HF: https://t.co/lkEGCMPL33

[Yoshi Suhara @suhara]

A new video game benchmark for LLM agents, designed across various game titles!

Happy to be part of this wonderful collaboration with @dongmin_park11 and the amazing team @Krafton_AI!

[Dongmin Park @dongmin_park11]

🚨New Paper Alert

As a game company, @Krafton_AI is actively exploring how to apply LLM agents to video games.

We present Orak—a foundational video gaming benchmark for LLM agents!

Includes Pokémon, StarCraft II, Slay the Spire, Darkest Dungeon, Ace Attorney, and more in🧵

[Yoshi Suhara @suhara]

Llama Nemotron Nano 4B matches the previous Nano 8B accuracy with only half the size!

🤗 https://t.co/V17tGSfu8U

[NVIDIA AI Developer @NVIDIAAIDev]

🤝 Meet NVIDIA Llama Nemotron Nano 4B, an open reasoning model that provides leading accuracy and compute efficiency across scientific tasks, coding, complex math, function calling, and instruction following for edge agents.

✨ Achieves higher accuracy and 50% higher throughput than other leading open models with 8 billion parameters or fewer
📗 Supports hybrid reasoning, optimized for low-cost inference
👨‍💻 Deploy at the edge with NVIDIA Jetson and NVIDIA RTX GPUs, maximizing security, flexibility, and cost

Now on @huggingface 📥 https://t.co/Pfyw99o8bv

[Oleksii Kuchaiev @kuchaev]

NeMo RL is now open source! It replaces NeMo-Aligner and is the toolkit we use to post train next generations of our models. Give it a try https://t.co/IZagZOhBK9