Hardware profiles

Hardware profiles

The server is published as two images. Match the image to your hardware.

NVIDIA x86_64 (CUDA cu128)

ghcr.io/forgeguard-ai/kokoro-server:latest (alias kokoro-server-cu128).

• Supported. Built on CUDA 12.8 with cuDNN. The cu128 PyTorch wheels carry kernels for compute capabilities sm_86 through sm_120 — that is NVIDIA RTX 3000 (Ampere) through RTX 5000 (Blackwell) in a single image, plus datacenter GPUs in that range.
• Runs as a container with --gpus all. Also runs on CPU with USE_GPU=false.

NVIDIA Jetson Orin (arm64, JetPack 6)

ghcr.io/forgeguard-ai/kokoro-server-jetson:latest.

• Supported on Jetson Orin (compute capability sm_87) running JetPack 6 / L4T r36 (CUDA 12.6, cuDNN 9.3). Confirm your L4T version with cat /etc/nv_tegra_release.

• Launch with the NVIDIA runtime:

  docker run -d --name kokoro --runtime nvidia -p 8880:8880 \
    ghcr.io/forgeguard-ai/kokoro-server-jetson:latest

• The Jetson image pins its CUDA math libraries to the JetPack 6 toolkit to avoid a cuBLAS version conflict; use the Jetson image (not the cu128 image) on these devices.

CPU

• Supported, reduced throughput. The cu128 image runs on CPU with USE_GPU=false. Expect substantially slower synthesis than on a GPU. Useful for CI, development, and low-volume use.

AMD (ROCm) and Intel

• Planned, not currently supported. Tracked on the roadmap; do not treat these as available today.

Compatibility at a glance

Target	Status	Notes
NVIDIA RTX 3000 → 5000 (x86_64, CUDA cu128)	Supported	Single cu128 image (sm_86–sm_120).
NVIDIA datacenter GPUs in that range	Supported	Same cu128 image.
NVIDIA Jetson Orin (arm64)	Supported	JetPack 6 / L4T r36; use the Jetson image.
CPU (x86_64)	Supported	USE_GPU=false; reduced throughput.
Apple Silicon (MPS)	Experimental	Auto-detected for local runs; not a published container target.
AMD (ROCm), Intel	Planned	Not currently supported.

See Compatibility for API and version details.