Models that fit 4 GB RAM

24 models fit a device with 4 GB of memory by the size rule of thumb. Pick iOS or Android below for a stricter usable-memory check. 5 of them fit by using their smaller Q3_K_M file. Download once, then use it offline with no account and no API fees.

Can my phone run this?

Operating system

Showing 3 of 3 models (filtered from 135)

DeepSeek-R1-Distill-Qwen-1.5B

by bartowski · built on DeepSeek R1 Distill Qwen 1.5B
#20
Official1.5BThinking
50Kdownloads95likes

An official 1.5B reasoning model based on DeepSeek R1 Distill Qwen 1.5B, which works through a problem step by step. The Q4_K_M file is 1.12 GB.

Run it with PocketPal (iOS, Android), ChatterUI (Android) or Layla (iOS, Android).

iOS-safe from 4 GB · Android-safe from 4 GB · 1.12 GB download · ~18–30 tok/s (estimate)

Gemma-3-1B-it-GLM-4.7-Flash-Heretic-Uncensored-Thinking

by Andycurrent · built on gemma 3 1b pt
#93
NewUncensored1BThinking
13Kdownloads169likes

Gemma-3-1B-it-GLM-4.7-Flash-Heretic-Uncensored-Thinking is a 1B model by Andycurrent of gemma 3 1b pt, using heretic, an automated abliteration tool. It writes out its reasoning before it answers. The Q4_K_M file is 0.81 GB.

Run it with PocketPal (iOS, Android), ChatterUI (Android) or Layla (iOS, Android).

iOS-safe from 4 GB · Android-safe from 4 GB · 0.81 GB download · ~25–45 tok/s (estimate)

Qwen3-1.7B-Magic_decensored

by MagicalAlchemist · built on Qwen3 1.7B Base
#62
Uncensored1.7BThinking
Uncensored IQ 32.3
Reasoning5.5
Answers7.5/10
289downloads0likes

Qwen3-1.7B-Magic_decensored is a 1.7B model by MagicalAlchemist of Qwen3 1.7B Base, using heretic, an automated abliteration tool. It writes out its reasoning before it answers. The Q4_K_M file is 1.11 GB.

Run it with PocketPal (iOS, Android), ChatterUI (Android) or Layla (iOS, Android).

iOS-safe from 4 GB · Android-safe from 4 GB · 1.11 GB download · ~18–30 tok/s (estimate)

No model matches. Try a higher memory setting, Anything, or clear the search.
Copied