Exxact Corporation

08/28/2026

New benchmark: LoRA fine-tuning on NVIDIA RTX PRO Blackwell GPUs. We measured multi-GPU throughput, DDP efficiency, power draw, and VRAM across RTX PRO 4500/5000/6000 SE. Full results + test setup inside. https://bit.ly/4xcxM9f

08/27/2026

Most say a 284B model can’t run on a single NVIDIA DGX Spark. We proved it can—and then compared Ollama, vLLM, and DS4 across quality, tokens/sec, concurrency, and memory stability. See which inference engine fits your workload. https://bit.ly/4ygRlOc

08/26/2026

The Exxact TensorEX 8U HGX B300 Server now ships on shorter lead times — 8x NVIDIA B300 GPUs, 1.8TB/s NVLink, built for trillion-param AI training. Inquire today: https://bit.ly/3SuD5Bi

08/25/2026

BioTuring builds AI platforms for single-cell and spatial biology research. GPU memory limits forced batching and capped sample sizes, slowing scientists down. With the help of Exxact deployed 3 NVIDIA DGX B300 systems, unlocking full-scale pipelines and faster AI. Read the case study. https://bit.ly/4xUWMBc

BioTuring NVIDIA

08/21/2026

Inferencing on-prem matters when latency and data control are non-negotiable. NVIDIA DGX Station™ is designed for serious AI development and deployment, helping you keep inference close to the work. If cloud constraints or unpredictable performance are a pain point, this helps. Secure your DGX Station through Exxact, initial shipments are underway: https://bit.ly/4xKvKMW

08/18/2026

Generative AI produces content from a prompt. Agentic AI goes further—planning, iterating, and using tools to complete goals autonomously. See the side-by-side differences and how each impacts hardware needs (GPU, CPU, RAM, NVMe, and I/O).
https://bit.ly/4fodjan

08/17/2026

Need to keep your data private and off the cloud? Exxact can help. Exxact is proud to offer NVIDIA RTX PRO Workstations, DGX Spark, and DGX Station to run the recently announced NVIDIA Nemotron 3.5 Lightning.

⚡ Fast & efficient AI agents don’t have to live in the cloud.

Meet NVIDIA Nemotron 3.5 Lightning: a new open, customizable 30B MoE model delivering up to 4X faster token generation and 30% faster time to completion compared with open models in its class.

Run it locally on NVIDIA RTX GPUs, DGX Spark and DGX Station today.

Learn more and explore the latest NVIDIA Local AI updates: https://blogs.nvidia.com/blog/local-ai-open-source-models-agents-nemotron/

08/17/2026

Fine-tuning is where models get practical. NVIDIA DGX Station™ is designed for serious on-prem AI development, so you can iterate quickly and keep work close to your data. If fine-tuning keeps getting delayed by limited access or slow cycles, this helps. Place your order with Exxact, shipping has begun: https://bit.ly/4gwFvZg

08/14/2026

Scaling to a 4× NVIDIA DGX Spark cluster changes the game: you’ll need a managed switch, the right QSFP cabling (standard or breakout), and a realistic power plan. We break down the build sheet + setup tips. https://bit.ly/4xFAmns

08/13/2026

Agentic AI works best when each step uses the model best suited to the job. NVIDIA Nemotron 3.5 Lightning and NeMo Switchyard give enterprises a flexible way to combine high-throughput specialized ex*****on with intelligent multi-model routing.

At Exxact, we help organizations bring these workflows into production with purpose-built, GPU-accelerated infrastructure for efficient local AI deployment.