MINIX has announced two new compact systems. The T4000 and T5000 Generative AI Mini Workstations designed for users who want to run AI workloads locally instead of relying on cloud services.
Both models focus on tasks like large language model (LLM) inference, generative content creation, and on-premise deployments, all within a relatively small desktop footprint.
Built on NVIDIA’s Blackwell Platform
At the core of these systems are modules based on NVIDIA’s Jetson AGX Thor platform, built on the newer Blackwell architecture.
Rather than traditional desktop GPU specs, performance here is measured in AI compute, with both models targeting local inference and parallel workloads.
Core AI Compute
| Model | AI Performance | GPU Cores | Tensor Cores | Additional Features |
|---|---|---|---|---|
| T4000 | Up to 1200 FP4 TFLOPs | 1536+ | 5th Gen | MIG, PVA 3.0 |
| T5000 | Up to 2070 FP4 TFLOPs | Up to 2560 | 5th Gen | MIG, PVA 3.0 |
In practical terms, this allows both systems to run 7B to 70B parameter models locally, depending on configuration and optimization.
CPU and Memory Configuration
Instead of focusing on traditional desktop CPU specs, both systems prioritize memory bandwidth and efficiency—key factors for AI workloads.
CPU and Memory
| Model | CPU | Memory | Bandwidth |
|---|---|---|---|
| T4000 | 12-core Arm Neoverse V3AE | Up to 128GB LPDDR5X | ~273 GB/s |
| T5000 | 14-core Arm Neoverse V3AE | Up to 128GB LPDDR5X | ~273 GB/s |
This setup is designed to handle large datasets, multi-modal AI, and concurrent processes without bottlenecks.
Connectivity and I/O
Despite their size, both systems are equipped more like compact servers than standard mini PCs.
Ports and Connectivity
| Feature | Specification |
|---|---|
| Ethernet | Dual 10GbE |
| Wireless | Wi-Fi 6E, Bluetooth 5.3 |
| Display | 2× HDMI 2.1 (4K @ 60Hz) |
| USB | 4× USB-A, 1× USB-C |
| Power | 24V DC, up to 200W |
This configuration makes them suitable for shared environments, edge deployments, or local AI hubs.
Storage and Physical Design
Both models come with fast storage and a compact chassis that doesn’t require rack mounting.
Storage and Build
| Feature | Specification |
|---|---|
| Storage | 1TB PCIe 4.0 SSD (up to 4TB) |
| Dimensions | 139.3 × 131 × 76.8 mm |
| Weight | ~1.42 kg |
| Cooling | Dual-fan system |
| Build | Metal + plastic chassis |
The design focuses on keeping the system compact while maintaining sustained performance under load.
Local AI and Software Support
One of the main points here is on-device AI processing. Instead of sending data to the cloud, everything can run locally.
The systems ship with Ubuntu 24.04 LTS and support NVIDIA’s full AI stack, including CUDA and TensorRT. This allows developers and teams to deploy AI workloads in a more controlled environment, especially for cases where privacy or latency is a concern.
Would you use/buy this?
MINIX is positioning these systems for professional workflows such as local AI assistants, generative media creation, enterprise AI deployments, and lightweight model training.
Rather than replacing traditional desktops, these devices are better understood as compact AI workstations or edge servers, built for specific workloads.
The T4000 and T5000 highlight a shift toward local AI computing, where performance is measured less by gaming or productivity benchmarks and more by how efficiently a system can run AI models.
What stands out here is the balance between form factor and capability. Systems like these bring workstation-level AI compute into a much smaller footprint, which could make them more practical for smaller teams or individual developers.
That said, this is still a niche category. Pricing and local availability will play a big role in determining whether setups like this become more common locally.
