text model · Laguna · Android
Can I run Laguna XS 2.1 on Google Pixel 10 Pro?
No. Laguna XS 2.1 needs ~22.2 GB even at Q4_K_M, but Google Pixel 10 Pro only has ~10.5 GB usable.
Needs ~22.2 GB even at Q4_K_M, but only ~10.5 GB is usable.
That figure is at a 4k context and moves about ±15% as context length changes.
The gap is about 11.7 GB: Laguna XS 2.1 needs roughly 22.2 GB at Q4_K_M and Google Pixel 10 Pro leaves only about 10.5 GB usable for a model. The lightest tracked hardware that runs Laguna XS 2.1 is the Nvidia GeForce RTX 4090 (24GB) at 24 GB. See Laguna XS 2.1 on Nvidia GeForce RTX 4090 (24GB).
- Q4_K_M needed
- ~22.2 GB
- Usable on device
- ~10.5 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
- 33.4B (MoE, 3B active)
- Q4_K_M size
- 20 GB
- Q8_0 size
- 36 GB
- Context
- 256k
- Ollama tag
- laguna-xs-2.1
- Memory
- 16 GB ram
- Usable for weights
- ~10.5 GB
- Power draw
- ~7 W
- Best runtime
- llama.cpp (PocketPal) or MLC-LLM (Adreno GPU path)
What you can run instead
Run Laguna XS 2.1 on other hardware
FAQ
Can Google Pixel 10 Pro run Laguna XS 2.1?
No. Laguna XS 2.1 needs ~22.2 GB even at Q4_K_M, but Google Pixel 10 Pro only has ~10.5 GB usable.
How much memory does Laguna XS 2.1 need?
Google Pixel 10 Pro does not have enough memory. At Q4_K_M the weights are ~20 GB; with KV cache and runtime overhead, budget ~22.2 GB at a 4k context. It is a Mixture-of-Experts model (33.4B total / 3B active), so all experts must stay in memory; memory tracks total params, not active params.
What is the best tool to run Laguna XS 2.1 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/laguna-xs-2.1/pixel-10-pro) 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.