text model · Muse · Android
Can I run Muse Glimmer 30B on Samsung Galaxy S26 Ultra (16GB, 1TB config)?
No. Muse Glimmer 30B needs ~18.2 GB even at Q4_K_M, but Samsung Galaxy S26 Ultra (16GB, 1TB config) only has ~12 GB usable.
Needs ~18.2 GB even at Q4_K_M, but only ~12 GB is usable.
That figure is at a 4k context and moves about ±15% as context length changes.
The gap is about 6.2 GB: Muse Glimmer 30B needs roughly 18.2 GB at Q4_K_M and Samsung Galaxy S26 Ultra (16GB, 1TB config) leaves only about 12 GB usable for a model. The lightest tracked hardware that runs Muse Glimmer 30B is the Nvidia GeForce RTX 4090 (24GB) at 24 GB. See Muse Glimmer 30B on Nvidia GeForce RTX 4090 (24GB).
- Q4_K_M needed
- ~18.2 GB
- Usable on device
- ~12 GB
- Device memory
- 16 GB
Which quant fits
How to run it
On Android use PocketPal AI (Polished app, download GGUF and run offline.).
- Parameters
- 30B
- Q4_K_M size
- 16.12 GB
- Q8_0 size
- 30.12 GB
- Context
- 128k
- Ollama tag
- muse-glimmer:30b
- Memory
- 16 GB ram
- Usable for weights
- ~12 GB
- Best runtime
- llama.cpp (PocketPal) or MLC-LLM (Adreno GPU path)
What you can run instead
Run Muse Glimmer 30B on other hardware
FAQ
Can Samsung Galaxy S26 Ultra (16GB, 1TB config) run Muse Glimmer 30B?
No. Muse Glimmer 30B needs ~18.2 GB even at Q4_K_M, but Samsung Galaxy S26 Ultra (16GB, 1TB config) only has ~12 GB usable.
How much memory does Muse Glimmer 30B need?
Samsung Galaxy S26 Ultra (16GB, 1TB config) does not have enough memory. At Q4_K_M the weights are ~16.12 GB; with KV cache and runtime overhead, budget ~18.2 GB at a 4k context.
What is the best tool to run Muse Glimmer 30B on Android?
On Android, PocketPal AI (Polished app, download GGUF and run offline.) is the go-to option. NPU acceleration is limited and chip-specific; most apps run on CPU. Expect 1B-4B class.
Embed this
A live badge for your model card or README, updated as the data is. [](https://localmodel.run/can-i-run/muse-glimmer-30b/samsung-s26-ultra) Sources
Weights are measured from GGUF files; KV cache and overhead are computed, so totals can vary ~15% with context and runtime. Any tok/s is a bandwidth estimate. See methodology.