[Blog] Generative AI is moving from "answering" to "acting." Starting with agent-type AI using NVIDIA Nemotron.

アプライド

With the spread of generative AI, including ChatGPT, AI has begun to be utilized in various scenarios such as text generation, summarization, translation, and programming assistance. On the other hand, in corporate and research and development settings, there is an increasing expectation for more advanced AI applications, not just to answer questions, but to: - Automatically search for necessary information from internal technical documents - Compare multiple pieces of information to determine the next course of action - Operate external tools and business systems - Progress through a series of tasks including research, analysis, code generation, and report creation This system, which infers based on objectives and utilizes necessary information and tools to advance tasks, is referred to as "Agentic AI." Supporting this Agentic AI is NVIDIA's open model family, known as NVIDIA Nemotron. In this article, we will introduce what NVIDIA Nemotron is, how it differs from traditional generative AI, and what applications can be considered in manufacturing and research and development.

Date and time
Capital
Entry fee

Related product information

AI Development, Local LLM, AI Workstation for DL
AI Development, Local LLM, Deep Learning AI Workstation

Recommended models for AI development, local LLM, and deep learning applications: Core Ultra processors, high-end GPUs like RTX 5090/RTX PRO 6000, and models equipped with Linux Ubuntu.
Rack-mounted servers for HPC, LLM, and generative AI development.
Rack-mounted servers for HPC, LLM, and generative AI development.

At Applied, we offer optimal computers tailored to your price and specifications through full custom orders starting from one unit. *For more details, please refer to the PDF materials or feel free to contact us.
We will introduce various HPCs for different applications of LLMs (Large Language Models).
Introducing models for LLM (Large Language Model) training and models for inference!

If you're looking for a model for LLM (Large Language Model) inference, here it is! APPLIED WST-XW93475XS3Q2TTNVM ■ CPU: Xeon W9-3475X ■ Memory: 256GB (32GB×8) DDR5-4800 ■ Storage: 2TB M.2 NVMe-SSD ■ GPU: [2GPU] NVIDIA RTX6000 Ada 48GB-GDDR6 ■ OS: Ubuntu 22.04 LTS ■ Power Supply: [2PSU] 1000W ■ 3-year send-back hardware warranty 5,998,000 yen (tax included) APPLIED WST-XW73465XS3Q2TTNVM ■ CPU: Xeon W7-3465X ■ Memory: 256GB (32GB×8) DDR5-4800 ■ Storage: 2TB M.2 NVMe-SSD ■ GPU: NVIDIA RTX6000 Ada 48GB-GDDR6 ■ OS: Ubuntu 22.04 LTS ■ Power Supply: [2PSU] 1000W ■ 3-year send-back hardware warranty 4,498,000 yen (tax included)
High-Performance High-End Servers for Generative AI/LLM [A Must-See for AI Developers]
Introducing the new rack-mounted high-performance high-end server, optimized for generative AI/LLM/deep learning development!

At Applied, we offer optimal computers tailored to your price and specifications through full custom orders starting from one unit. *For more details, please refer to the PDF materials or feel free to contact us.
A small AI supercomputer that will revolutionize local AI development.
A compact AI supercomputer "based on NVIDIA DGX Spark" that will revolutionize local AI development.

- Computational Performance: Equipped with NVIDIA GB10 GraceBlackwell Superchip (CUDA 6144 cores + 5th generation Tensor cores), offering up to 1,000 TOPS of FP4 AI performance - CPU: 20-core ARM Grace (Cortex-X925×10 + A725×10) - Memory: 128GB LPDDR5X (coherent sharing between CPU and GPU) - Storage: 1TB or 4TB NVMe M.2 SSD (self-encrypting support) - Size/Weight: Approximately 151×151×52mm, weight about 1.2kg - Connectivity: USB 3.2 Type-C ×4, 10GbE RJ-45, Wi-Fi 7, Bluetooth 5.3 - Power/Cooling: Maximum power consumption approximately 224W, includes 240W PSU, compact chassis cooling design - OS/Framework: Compatible with NVIDIA DGX OS, pre-installed environment for AI development such as Isaac/Holoscan
GPU product lineup supporting generative AI and HPC.
From generative AI, LLM, CAE to 3DCG production. GPU solutions supporting next-generation research and development.

We offer a lineup of NVIDIA GPUs optimized for generative AI, LLM, CAE, video production, and scientific computing. The "RTX PRO Blackwell series" for workstations accelerates large-scale AI inference, 3DCG, CAD, and simulation processing with up to 96GB GDDR7 VRAM and PCIe 5.0 support. The "H200 NVL" for servers features 141GB HBM3e memory and an ultra-fast bandwidth of 4.8TB/s, catering to large-scale LLM and HPC applications. Additionally, we also feature the compact AI supercomputer "DGX Spark" and the cost-effective "GeForce RTX 50 series." ■ Key Specifications - Up to 141GB HBM3e / 96GB GDDR7 VRAM - PCIe 5.0 support - Equipped with CUDA / Tensor / RT cores - High bandwidth design for AI inference, CAE, and 3DCG - Models available for both workstations and servers
NVIDIA Alpamayo compatible HPC workstation
HPC infrastructure to accelerate the inference and simulation of autonomous driving AI.

CPU: AMD Ryzen Threadripper PRO 9965WX (24 cores / 48 threads, 4.2–5.4GHz, 350W) Memory: 128GB (32GB x 4) DDR5-5600 Registered ECC DIMM Storage: 2TB M.2 NVMe PCIe 4.0 GPU: NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition 96GB GDDR7 OS: Ubuntu 24.04 LTS installation service + framework environment setup Power Supply: 1200W / 100V, 80PLUS Platinum certified, ATX3.1 Warranty: 3-year send-back hardware warranty We also support customization according to your needs, such as memory and storage upgrades, GPU changes, OS changes, and parts configuration focused on cooling and durability.
NVIDIA H200 NVL 4GPU AI server
H200-equipped server that accelerates large-scale AI/HPC with NVLink 4-Way.

【Xeon 6 Model】 OS: Linux/Windows CPU: Intel Xeon 6730P×2 Memory: 512GB ECC R-DIMM Storage: U.2 NVMe SSD 1.92TB GPU: NVIDIA H200 NVL 141GB×4, NVLink Bridge Board 4-way Passive×2 Power Supply: Redundant 3+1 3,200W/200V 3-year send-back warranty. 【EPYC Model】 OS: Linux CPU: AMD EPYC 9455 Memory: 512GB ECC R-DIMM Storage: U.2 NVMe SSD 1.92TB GPU: NVIDIA H200 NVL 141GB×4, same NVLink configuration Power Supply: Redundant 3+1 3,200W/200V 3-year send-back warranty.
NVIDIA PhysicsNeMo compatible GPU workstation
A GPU environment for PhysicsNeMo that integrates physical laws and AI to accelerate design verification.

**Ryzen 9 Model** CPU: AMD Ryzen 9 9950X3D (16 cores / 32 threads) Memory: 128GB DDR5-5600 SSD: 2TB M.2 NVMe GPU: NVIDIA RTX PRO 4500 Blackwell 32GB OS: Ubuntu 24.04 LTS Power Supply: 850W 80PLUS GOLD 1-year send-back warranty **Threadripper Model** CPU: AMD Ryzen Threadripper 9970X (32 cores / 64 threads) Memory: 256GB DDR5-5600 ECC RDIMM SSD: 4TB M.2 NVMe GPU: NVIDIA RTX PRO 6000 Blackwell Max-Q 96GB OS: Ubuntu 24.04 LTS Power Supply: 1200W 80PLUS PLATINUM 3-year send-back warranty Customization is available to match your usage and analysis scale.
Applied PC equipped with NVIDIA RTX Pro 6000
NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition equipped

NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition model Be-Clia Workstation Type-ZU2V2-9 ■ OS: Windows 11 Pro 64bit ■ CPU: Intel Core Ultra9 285K (Pcore: 3.7-5.5GHz / Ecore: 3.2-4.6GHz / 24C (P8+E16) / 24T) ■ Memory: 128GB (64GB ×2) DDR5-6400 ■ SSD: 2TB (M.2 NVMe) ■ GPU: NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition 96GB GDDR7 ECC ■ Power Supply: 1000W / 100V 80PLUS Gold certified (ATX3.1) ■ Warranty: 1-year send-back hardware warranty