# Can I run Kokoro-82M on Apple M5 Pro (48GB)?

Updated: 2026-06-15

**Yes, it runs.** Runs at fp32 using ~1 GB of ~32 GB usable.

- Model: 82M Text to speech, peak ~1 GB memory at fp32
- Device: 48 GB unified, ~32 GB usable
- Tools: kokoro (Python), ONNX Runtime, Kokoro-FastAPI, CoreML
- License: Apache-2.0 (commercial use: yes)

Sources: https://huggingface.co/hexgrad/Kokoro-82M, https://huggingface.co/FluidInference/kokoro-82m-coreml, https://github.com/remsky/Kokoro-FastAPI, https://github.com/puff-dayo/Kokoro-82M-Android.

More: https://localmodel.run/can-i-run/kokoro-82m/apple-m5-pro-48gb