Hardware

One small machine you own.

The vault runs 24/7 on hardware you purchase and own. Here is what to buy and what it costs to run.

Recommended hardware

ADAMAS runs entirely on a machine you own. It is hardware-flexible: anything that can run the local Hermes model and stay powered on 24/7 will work. Below are three configurations we have tested, from a quiet default to a large-model option. These are examples, not requirements.

Apple Mac mini (M4), 24–32 GB — recommended for most teams

  • Quietest and most power-efficient option; near-silent, a few dollars a month in electricity.
  • Runs Hermes comfortably for capture, evaluation, and asset generation at the model sizes ADAMAS uses.
  • Simplest to set up and maintain — no driver management.
  • Approx. 2026 price: ~$600–$1,000, depending on memory and storage.

AMD Ryzen AI Max+ 395 mini PC (e.g. GMKtec EVO-X2), 128 GB — more local headroom

  • 128 GB of unified memory lets Hermes run larger 70B-class models entirely on-device for higher-fidelity assets.
  • x86/Linux; integrates with the same local pipeline.
  • Slightly more initial configuration (handled during your build); fan audible but not intrusive under load.
  • Approx. 2026 price: ~$1,800–$2,000.

NVIDIA DGX Spark (GB10), 128 GB — optional, for CUDA/NVIDIA environments

  • 128 GB unified memory with the full CUDA stack; strong for prototyping and on-device fine-tuning.
  • Premium price; for ADAMAS's core capture-and-generate workload it is more than most teams need, and large-model token speed is bandwidth-limited.
  • Approx. 2026 price: ~$3,000–$4,000.

All three run ADAMAS fully offline, expose no inbound ports, and keep the ledger, source material, embeddings, and generated assets on the machine. Optional hybrid-cloud tasks remain opt-in and are logged. A typical local-first setup replaces roughly $200–459/month in cloud-AI subscriptions (run your own numbers).

Hardware names are provided only as examples of configurations we have tested. They are not requirements, and ADAMAS is hardware-flexible. Apple and Mac mini are trademarks of Apple Inc.; NVIDIA and DGX Spark are trademarks of NVIDIA Corporation; AMD and Ryzen are trademarks of Advanced Micro Devices, Inc.; other names are the property of their respective owners. ADAMAS and Falcon Intelligence Group are not affiliated with, endorsed by, sponsored by, or partnered with Apple, NVIDIA, AMD, GMKtec, or any hardware manufacturer. Performance depends on configuration, model, and workload; figures are indicative and not guaranteed.