NVIDIA Nemotron 3 Super
Nemotron 3 Super is an NVIDIA open-weight language model released on March 11, 2026, with 120B total and 12B active parameters, optimized for collaborative agents and high-volume workloads. It produces a reasoning trace before its answer and supports up to 1M tokens of context.
Released
March 11, 2026
Type
reasoning
License
open-weight
NVIDIA Nemotron Open Model License
Context
1,048,576 tokens
NVIDIA lists up to 1M tokens (1,048,576 as the maximum model length); its serving recipes default to 256K.
Capabilities
Architecture
Parameters: 120B total, 12B active (LatentMoE: Mamba-2 + MoE + attention hybrid)
Sources
Every figure on this page was checked against these pages on . A figure we could not source is left out.
- Model card: nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16 (opens in a new tab)
- NVIDIA Nemotron 3 Super Technical Report (opens in a new tab)
- NVIDIA Developer: Nemotron (opens in a new tab)
Spotted something out of date? Tell us.
Links
In the news: NVIDIA Nemotron 3 Super and NVIDIA
MarkTechPost
Liquid AI Releases LFM2.5-24B-A2B: A 2B-Active MoE for On-Device Agents
Liquid AI has released LFM2.5-24B-A2B, an open-weight reasoning model for agents designed to run on laptops and PCs. This 24B-parameter Mixture-of-Experts (MoE) model activates about 2B parameters per token and is the final member of the LFM2.5 family. The model is trained within popular agent harnesses, focusing on multi-step tool use rather than open chat. Base and post-trained weights are available on Hugging Face. LFM2.5-24B-A2B features a 131,072-token context window and can run on laptops and mini PCs using Apple M5 Max or AMD Ryzen AI Max+ 395 processors via GGUF/MLX. It also supports single NVIDIA B300 for serving and offers FP8, GGUF, ONNX, and MLX 4/5/6/8-bit builds. In performance tests, LFM2.5-24B-A2B leads or matches larger MoEs on five out of eight agentic and tool-use benchmarks. It achieved 27.69 on AutomationBench, more than triple the score of Qwen3.6-35B-A3B. However, it performed less strongly on SWE-Bench Verified (OpenCode) with a score of 51.00 compared to Qwen3.6-35B-A3B's 66.20. The model is text-only and reasoning-only, built for local agent loops, with each assistant turn beginning with a <think> block. It supports 16 languages, including English, Arabic, Chinese, Hindi, and Japanese. Its architecture retains LFM2’s 40-layer hybrid structure, incorporating 30 double-gated short convolution blocks and 10 grouped query attention (GQA) blocks. Each MoE block contains 64 experts with top-4 routing. Context increased from 32,768 to 131,072 tokens, and vocabulary doubled from 65K to 128K via an in-place tokenizer expansion for improved non-Latin tokenization. Training involved scaling pre-training from 17T to 33T tokens, followed by mid-training with 2T tokens at a 32K context and then 300B tokens at a 131K context, focusing on tool calls and multi-turn interactions.
NVIDIA Newsroom
NVIDIA Commits $250 Million to Advance UK Agentic Safety and Security Research Over the Next Five Years
NVIDIA announced a five-year compute donation valued at $250 million to strengthen the United Kingdom’s agentic safety and security research capability. This commitment aims to advance research in AI safety and security within the UK.
NVIDIA Newsroom
European Cities Deploy NVIDIA-Powered AI Agents to Run Smarter Urban Operations
European cities are deploying NVIDIA-powered AI agents and digital twins to enhance urban operations. These AI systems are designed to assist city teams in making daily decisions, such as scheduling road repairs and improving energy efficiency. The initiative utilizes NVIDIA's accelerated computing platform to build these advanced physical AI agents and digital twins. This technology aims to create smarter urban environments by optimizing resource allocation and operational management across the EMEA region.
NVIDIA Newsroom
OCP Global Summit: Amid Ecosystem Adoption, NVIDIA Opens Vera Rubin Architecture to Help Scale AI Factories Worldwide
At the Open Compute Project (OCP) Global Summit, NVIDIA announced the Vera Rubin architecture, designed to help scale AI factories globally as they expand from megawatts to gigawatts. The company stated that open standards are crucial for the industry to scale AI infrastructure more rapidly, efficiently, and reliably. NVIDIA also revealed new partnerships and contributions at the summit, which is being held at the San Jose McEnery Convention Center. These initiatives aim to further support the growth and development of AI infrastructure.
Cisco Blogs
Cisco and NVIDIA Announce USFP: Expanding the Path to AI Connectivity
Cisco and NVIDIA announced USFP, the first 400G-per-lane pluggable form factor. This new technology is designed to enable efficient scaling of AI infrastructure. The companies state it maintains proven operations.