Hardware Requirements

Mac and Linux hardware requirements for running BodhAI.

Hardware Requirements


Mac (development and small teams)

BodhAI on Mac uses the MLX framework for Metal GPU acceleration on Apple Silicon.

Model size Minimum RAM Recommended
7B 16 GB M2 Pro or later
14B 24 GB M3 Pro / M4 Pro
32B 48 GB M3 Max / M4 Max

Required:

  • Apple Silicon (M1 or later) — Intel Mac is not supported
  • macOS 13 Ventura or later
  • 50 GB free disk space (model + document storage)

Recommended:

  • M3 Pro or later for teams of 3+ concurrent users
  • External SSD for document storage on large deployments

Linux (production deployments)

BodhAI on Linux uses llama-server (llama.cpp) with NVIDIA GPU acceleration (CUDA).

Model size Minimum VRAM GPU examples
7B 16 GB RTX 3090, A4000
14B 20 GB RTX 3090 Ti, A5000
32B 40 GB+ A100, 2× RTX 3090

Required:

  • NVIDIA GPU with CUDA 12.1+
  • NVIDIA Driver 525+
  • Docker + NVIDIA Container Toolkit
  • Ubuntu 22.04 or later (RHEL 8+ also supported)
  • 100 GB free disk space

Multi-GPU: llama-server supports splitting model layers across multiple GPUs. The 32B model can run across 2× 24 GB GPUs.


CPU-only fallback (Mac)

If mlx-lm fails to start, BodhAI falls back to llama-server (llama.cpp) on CPU. This works but is significantly slower — expect 2–5 tokens/second vs 35–55 on Metal GPU. Not recommended for team use.


Network

No special network requirements for local use. For remote access via Cloudflare Tunnel, the server needs outbound internet access on port 443.