audio model · kokoro · Android
Can I run Kokoro-82M on Samsung Galaxy S25 Ultra (16GB, 1TB config only)?
Yes. Kokoro-82M runs on Samsung Galaxy S25 Ultra (16GB, 1TB config only) at fp32 (~1 GB of ~12 GB usable).
Runs at fp32 using ~1 GB of ~12 GB usable.
- Peak memory
- ~1 GB
- Usable on device
- ~12 GB
- Device memory
- 16 GB
- Quant
- fp32
How to run it
Use kokoro (Python) or ONNX Runtime at fp32. It is light enough to run on CPU; a GPU just makes it faster.
- Type
- Text to speech
- Parameters
- 82M
- Peak memory
- ~1 GB at fp32
- License
- Apache-2.0
- Memory
- 16 GB ram
- Usable for weights
- ~12 GB
- Power draw
- ~8 W
- Best runtime
- llama.cpp (PocketPal) or MLC-LLM (Adreno GPU path)
You could also run
Run Kokoro-82M on other hardware
FAQ
Can Samsung Galaxy S25 Ultra (16GB, 1TB config only) run Kokoro-82M?
Yes. Kokoro-82M runs on Samsung Galaxy S25 Ultra (16GB, 1TB config only) at fp32 (~1 GB of ~12 GB usable).
How much memory does Kokoro-82M need?
Samsung Galaxy S25 Ultra (16GB, 1TB config only) has room to spare. At fp32 the realistic peak is ~1 GB of memory.
What do I use to run Kokoro-82M locally?
Kokoro-82M runs in kokoro (Python) or ONNX Runtime (among others). It runs on CPU, so no GPU is required.
Sources
- gadgetversus.com
- github.com/a-ghorbani
- github.com/google-ai-edge
- github.com/mlc-ai
- github.com/puff-dayo
- github.com/remsky
- github.com/shubham0204
- github.com/Vali-98
- gsmarena.com/samsung_galaxy_s25_ultra-12750.php
- gsmarena.com/samsung_galaxy_s25_ultra-13322.php
- huggingface.co/FluidInference
- huggingface.co/hexgrad
- layla-network.ai
- sammyfans.com
- sammyguru.com
- samsung.com
VRAM figures are sourced peak-usage anchors at the noted quant, validated 2026-08-03. See methodology.