text model · Kimi · Android
Can I run Kimi K2.7 Code on Samsung Galaxy S26 Ultra (16GB, 1TB config)?
No. Kimi K2.7 Code needs ~592.1 GB even at Q4_K_M, but Samsung Galaxy S26 Ultra (16GB, 1TB config) only has ~12 GB usable.
Needs ~592.1 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 580.1 GB: Kimi K2.7 Code needs roughly 592.1 GB at Q4_K_M and Samsung Galaxy S26 Ultra (16GB, 1TB config) leaves only about 12 GB usable for a model. No single tracked device has enough memory; Kimi K2.7 Code needs a multi-GPU or high-memory rig.
- Q4_K_M needed
- ~592.1 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
- 1000B (MoE, 32B active)
- Q4_K_M size
- 583.71 GB
- Context
- 256k
- 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
FAQ
Can Samsung Galaxy S26 Ultra (16GB, 1TB config) run Kimi K2.7 Code?
No. Kimi K2.7 Code needs ~592.1 GB even at Q4_K_M, but Samsung Galaxy S26 Ultra (16GB, 1TB config) only has ~12 GB usable.
How much memory does Kimi K2.7 Code need?
Samsung Galaxy S26 Ultra (16GB, 1TB config) does not have enough memory. At Q4_K_M the weights are ~583.71 GB; with KV cache and runtime overhead, budget ~592.1 GB at a 4k context. It is a Mixture-of-Experts model (1000B total / 32B active), so all experts must stay in memory; memory tracks total params, not active params.
What is the best tool to run Kimi K2.7 Code 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/kimi-k2.7-code/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.