Hong Kong-based MINIX has unveiled the latest product – the T4000 and T5000 Generative AI Mini Workstations – which are a couple of compact yet powerful machines built for professional generative AI workloads and local large language model (LLM) inference.
These new offerings are powered by NVIDIA Blackwell-based Jetson AGX Thor series modules with up to 2070 Sprase FP4 TFLOPs of processing power, up to 2560 GPU cores and 5th gen Tensor Cores. Mixing in Multi-Instance GPU (MIG) support, they are capable of parallel processing with hyper efficiency, as they can natively tackle low-latency local inference for LLMs ranging from 7B to 70B parameters.
To support large model operations, the T4000 and T5000 utilize high-core-count Arm Neoverse-V3AE CPUs (12 cores for T4000, 14 cores for T5000) and feature up to 128 GB of high-bandwidth LPDDR5X memory.
The mini workstations are built for robust professional use, featuring dual 10GbE Ethernet for fast data transfer, Wi-Fi 6E, and dual HDMI 2.1 display support alongside massive storage options – starting with a 1 TB PCIe 4.0 NVMe M.2 SSD – and are housed in a compact, durable chassis that requires no rack and external cooling.
Made to support local AI processing that ensures all data and models remain on-device, they come pre-installed with Ubuntu 24.04 LTS (NVIDIA JetPack 7.1) and support the full NVIDIA software ecosystem, including CUDA and TensorRT, making them ideal for private AI hubs, image/video creation, and lightweight model fine-tuning.
Like our Facebook Page here at NasiLemakTech.com for more news and in-depth reviews! Also, join our Facebook Group for insightful information and memes!
Subscribe to our YouTube channel too!
