Platforms

Supported platforms

The same capabilities run on different chip architectures. The usage above is identical; what differs is how fully each kind of hardware is used underneath.

x86
Desktops, workstations and small servers. Faster with a dedicated GPU.
Configurations we've testedOne GPU with 32 GB VRAM · 128 GB RAMTwo GPUs, 32 GB + 24 GB VRAM
ARM
Macs with Apple silicon, and ARM-based edge devices.
Configurations we've testedEdge AI device with 128 GB unified memory
RISC-VPreview
For the next generation of local AI chips.
Configurations we've testedIn progress; no tested configuration yet
PreviewRISC-V support is still in progress; available features and performance will fill in over successive releases.
Only machines we have actually run it on are listed; minimum specs are still being tested and will be added when done.
Under the hood

Mature open-source engines.

Built on llama.cpp, ONNX Runtime and MLX, with the backend and quantization chosen per chip. No manual configuration needed.

Don’t see your device?

Tell us.

We’re adapting model by model. Tell us your device and what you’d like to do with it.

Tell us