Privacy First • Zero Cloud Dependency

Nexus Core

Run Llama 3, Mistral, and Mixtral at lightning speeds. Your data never leaves your desk.

Configure Yours

Technical Specifications

Compute EngineDual NVIDIA RTX 4090 (48GB VRAM Total)
ProcessorAMD Ryzen Threadripper PRO 5975WX
Memory128GB DDR5 ECC RAM
Storage4TB NVMe Gen5 SSD (Read/Write up to 10,000 MB/s)
ConnectivityDual 10GbE SFP+, WiFi 6E, Bluetooth 5.3
OS SupportUbuntu LTS, Windows Pro, macOS (via Docker)

Choose Your Power

Lite Node

$2,499 / unit

  • 16GB VRAM (RTX 4080 Super)
  • Supports 7B - 13B models
  • Silent Cooling Design
  • Standard Support
Buy Lite

Enterprise Rack

$14,500 / unit

  • 96GB VRAM (A6000 Ada)
  • Supports massive 120B+ models
  • Hot-swappable Drives
  • 24/7 Dedicated Engineer
Contact Sales

Frequently Asked Questions

No. Once the initial model weights are downloaded, Nexus Core operates entirely offline. This ensures absolute privacy for your prompts and data.

Nexus Core is compatible with any GGUF, EXL2, or AWQ quantized model. This includes Llama 3, Mistral, Mixtral, Command R, and many more via Ollama or LM Studio.

Yes. It comes pre-configured with a web-based dashboard. Just plug in power and Ethernet, and you can start chatting via your browser immediately.