text model · dots.llm · Android
Can I run dots.llm1 on Samsung Galaxy S26 Ultra (16GB, 1TB config)?
No. dots.llm1 needs ~98.1 GB even at Q4_K_M, but Samsung Galaxy S26 Ultra (16GB, 1TB config) only has ~12 GB usable.
Needs ~98.1 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 86.1 GB: dots.llm1 needs roughly 98.1 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 dots.llm1 is the Apple M3 Ultra (256GB) at 256 GB. See dots.llm1 on Apple M3 Ultra (256GB).
- Q4_K_M needed
- ~98.1 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
- 142B (MoE, 14B active)
- Q4_K_M size
- 94.4 GB
- Q8_0 size
- 151.9 GB
- Context
- 32k
- 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 dots.llm1 on other hardware
FAQ
Can Samsung Galaxy S26 Ultra (16GB, 1TB config) run dots.llm1?
No. dots.llm1 needs ~98.1 GB even at Q4_K_M, but Samsung Galaxy S26 Ultra (16GB, 1TB config) only has ~12 GB usable.
How much memory does dots.llm1 need?
Samsung Galaxy S26 Ultra (16GB, 1TB config) does not have enough memory. At Q4_K_M the weights are ~94.4 GB; with KV cache and runtime overhead, budget ~98.1 GB at a 4k context. It is a Mixture-of-Experts model (142B total / 14B active), so all experts must stay in memory; memory tracks total params, not active params.
What is the best tool to run dots.llm1 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/dots-llm1/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.