Hardware Sovereignty

Supported Platforms & Hardware Compatibility

FineTuneMyAI adapts to your hardware. No cloud lock-in. Compile and execute low-rank matrix updates directly on your workstation GPU or unified memory pool.

Architecture

macOS (Apple Silicon)

ProcessorsApple M1, M2, M3, M4 (Pro, Max, Ultra)

Backend FrameworkApple MLX & Metal Performance Shaders (MPS)

Minimum Memory16 GB Unified Memory

Capabilities

Native 4-bit QLoRA via mlx-lm

Zero PCIe transfer bottleneck

Unified memory for models up to 70B

$ curl -fsSL https://finetunemyai.com/install.sh | bash

Architecture

Linux (NVIDIA CUDA)

ProcessorsRTX 3060, 3090, 4070, 4090, A100, H100

Backend FrameworkPyTorch >= 2.2 + CUDA >= 12.1 + bitsandbytes

Minimum Memory12 GB VRAM + 16 GB System RAM

Capabilities

bitsandbytes NF4 quantization

Paged AdamW optimizer

FlashAttention-2 support

$ curl -fsSL https://finetunemyai.com/install-linux.sh | bash

Architecture

Windows 11 (WSL2 / DirectML)

ProcessorsNVIDIA RTX 3000 & 4000 Series

Backend FrameworkWSL2 with NVIDIA Container Toolkit or DirectML

Minimum Memory12 GB VRAM + 16 GB System RAM

Capabilities

Full CUDA parity under WSL2

Local Ollama integration

Opaque node credential pairing

$ irm https://finetunemyai.com/install.ps1 | iex

Check your workstation readiness

Open the Devices console to run the automated 1-click hardware capability detector.

Detect Local Hardware