The Event Horizon of Local AI
Ultimate open-source AI machine — run the largest models at full speed.
Multi-GPU scaling, batch processing, fine-tuning, private inference at scale.
Price Range
$3,500 – $5,000
80–200
tokens / sec
Pooled GPU VRAM
70B–405B
model params
600W–1.2kW
full load
Built for: researchers • ML engineers • small businesses • content agencies • anyone wanting multi-GPU power on a proven Intel Xeon server chassis
Signature Capabilities
- Up to 3× NVIDIA GPUs (8–24 GB VRAM each)
- 70B–405B parameter models at usable speeds
- Massive parallel workloads & batch generation
- Fine-tuning, multi-user serving, research-grade performance
- Long-term cheaper than cloud API usage
Core Specifications
| CPU | Intel Xeon (multi-core · server grade) |
|---|---|
| RAM | 16–64 GB ECC DDR4 |
| Storage | 512 GB – 2 TB NVMe SSD |
| GPUs | 1–3× Nvidia RTX 2060 or better (8–24 GB VRAM each, configurable) |
| Max VRAM | Up to 72 GB pooled (3×24 GB) |
| Inference speed | ~80–200 tok/s (pooled GPU) |
| Power draw | 300–800 W under full load |
| OS | Linux (Ubuntu pre-configured) |