Free to use · 32 official models · 135 in total · stats updated Nov 19, 2026

Run AI on your phone, for free

Download a model once and run it on your own device. No subscription, no API fees, no account. It works without a network, and nothing you type leaves the phone.

32official models
135models in the directory
$0no subscription or API fees
Offlineafter one download

Browse models

Can my phone run this?

Operating system
Device memory
Selected memoryAny

Showing 45 of 45 models (filtered from 135)

gemma-4-12b-it

by unsloth · built on gemma 4 12B
#1
NewOfficial12BVision
2.6Mdownloads977likes

An official 12B vision model based on gemma 4 12B, made to understand images as well as text. The Q4_K_M file is 7.12 GB.

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

iOS-safe from 12 GB · Android-safe from 12 GB · 7.12 GB download · ~6–12 tok/s (estimate)

Uncensored9BVision
2.2Mdownloads1.9Klikes

Qwen3.5-9B-The-Defiant-Fable-Uncensored-Heretic-NEO-IMATRIX-MAX-MTP is a 9B model by DavidAU of Qwen3.5 9B, using heretic, an automated abliteration tool. It can read images and writes out its reasoning before it answers. The Q4_K_M file is 6.83 GB.

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

iOS-safe from 12 GB · Android-safe from 12 GB · 6.83 GB download · ~6–12 tok/s (estimate)

Qwen3.5-9B

by unsloth · built on Qwen3.5 9B
#2
NewOfficial9BVision
2Mdownloads1.1Klikes

An official 9B vision model based on Qwen3.5 9B, made to understand images as well as text. The Q4_K_M file is 5.68 GB.

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

iOS-safe from 12 GB · Android-safe from 8 GB · 5.68 GB download · ~6–12 tok/s (estimate)

#67
Uncensored9BVision
1.7Mdownloads287likes

Ornith-1.5-9B-uncensored is a 9B model by junafinity, using abliteration. It can look at a photo or screenshot and answer questions about it. The Q4_K_M file is 5.63 GB.

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

iOS-safe from 12 GB · Android-safe from 8 GB · 5.63 GB download · ~6–12 tok/s (estimate)

Qwen3.5-4B

by unsloth · built on Qwen3.5 4B Base
#3
Official4BVision
1.6Mdownloads578likes

An official 4B vision model based on Qwen3.5 4B Base, made to understand images as well as text. The Q4_K_M file is 2.74 GB.

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

iOS-safe from 6 GB · Android-safe from 4 GB · 2.74 GB download · ~12–22 tok/s (estimate)

gemma-4-E2B-it

by unsloth · built on gemma 4 E2B
#6
Official2BVision
814Kdownloads361likes

An official 2B vision model based on gemma 4 E2B, made to understand images as well as text. The Q4_K_M file is 3.11 GB.

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

iOS-safe from 6 GB · Android-safe from 4 GB · 3.11 GB download · ~18–30 tok/s (estimate)

Uncensored9BVision
801Kdownloads112likes

DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic is a 9B model by xtools-at, using heretic, an automated abliteration tool. It can read images and writes out its reasoning before it answers. The Q4_K_M file is 5.78 GB.

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

iOS-safe from 12 GB · Android-safe from 8 GB · 5.78 GB download · ~6–12 tok/s (estimate)

gemma-4-E4B-it

by unsloth · built on gemma 4 E4B
#7
Official4BVision
744Kdownloads699likes

An official 4B vision model based on gemma 4 E4B, made to understand images as well as text. The Q4_K_M file is 4.98 GB.

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

iOS-safe from 8 GB · Android-safe from 8 GB · 4.98 GB download · ~12–22 tok/s (estimate)

Qwen3.5-9B-Uncensored-HauhauCS-Aggressive

by HauhauCS · built on Qwen3.5 9B
#68
Uncensored9BVision
646Kdownloads2.6Klikes

Qwen3.5-9B-Uncensored-HauhauCS-Aggressive is a 9B model by HauhauCS of Qwen3.5 9B, using community fine-tuning. It can look at a photo or screenshot and answer questions about it. Languages noted on the card include Chinese. The Q4_K_M file is 5.63 GB.

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

iOS-safe from 12 GB · Android-safe from 12 GB · 5.63 GB download · ~6–12 tok/s (estimate)

Gemma-4-E4B-Uncensored-HauhauCS-Aggressive

by HauhauCS · built on Gemma 4 E4B
#69
Uncensored8B · 4.5B activeVision
597Kdownloads1.6Klikes

Gemma-4-E4B-Uncensored-HauhauCS-Aggressive is a 8B · 4.5B active model by HauhauCS of Gemma 4 E4B, using fine-tuning. It can look at a photo or screenshot and answer questions about it. The Q4_K_M file is 5.34 GB.

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

iOS-safe from 12 GB · Android-safe from 8 GB · 5.34 GB download · ~6–12 tok/s (estimate)

Huihui-Qwen3.5-9B-abliterated

by huihui-ai · built on Qwen3.5 9B
#4
Uncensored9BThinkingVision
Uncensored IQ 43.2
Reasoning19.8
Answers9.5/10
478Kdownloads84likes

Huihui-Qwen3.5-9B-abliterated is a 9B model by huihui-ai of Qwen3.5 9B, using abliteration. It can read images and writes out its reasoning before it answers. The Q4_K_M file is 5.63 GB.

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

iOS-safe from 12 GB · Android-safe from 8 GB · 5.63 GB download · ~6–12 tok/s (estimate)

Qwen3.5-4B-Uncensored-HauhauCS-Aggressive

by HauhauCS · built on Qwen3.5 4B
#71
Uncensored4BVision
353Kdownloads880likes

Qwen3.5-4B-Uncensored-HauhauCS-Aggressive is a 4B model by HauhauCS of Qwen3.5 4B, using uncensoring. It can look at a photo or screenshot and answer questions about it. Languages noted on the card include Chinese. The Q4_K_M file is 2.71 GB.

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

iOS-safe from 6 GB · Android-safe from 6 GB · 2.71 GB download · ~12–22 tok/s (estimate)

Qwen3-VL-4B-Instruct

by unsloth · built on Qwen3 VL 4B
#10
Official4BVision
274Kdownloads80likes

An official 4B vision model based on Qwen3 VL 4B, made to understand images as well as text. The Q4_K_M file is 2.5 GB.

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

iOS-safe from 4 GB · Android-safe from 4 GB · 2.5 GB download · ~12–22 tok/s (estimate)

Qwen3-VL-8B-Instruct

by unsloth · built on Qwen3 VL 8B
#14
Official8BVision
197Kdownloads103likes

An official 8B vision model based on Qwen3 VL 8B, made to understand images as well as text. The Q4_K_M file is 5.03 GB.

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

iOS-safe from 8 GB · Android-safe from 8 GB · 5.03 GB download · ~6–12 tok/s (estimate)

gemma-3-4b-it

by unsloth · built on gemma 3 4b pt
#16
Official4BVision
178Kdownloads217likes

An official 4B vision model based on gemma 3 4b pt, made to understand images as well as text. The Q4_K_M file is 2.49 GB.

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

iOS-safe from 4 GB · Android-safe from 4 GB · 2.49 GB download · ~12–22 tok/s (estimate)

Huihui-Qwen3.5-4B-abliterated

by huihui-ai · built on Qwen3.5 4B
#72
Uncensored4BVision
158Kdownloads35likes

Huihui-Qwen3.5-4B-abliterated is a 4B model by huihui-ai of Qwen3.5 4B, using abliteration. It can look at a photo or screenshot and answer questions about it. The Q4_K_M file is 2.71 GB.

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

iOS-safe from 4 GB · Android-safe from 4 GB · 2.71 GB download · ~12–22 tok/s (estimate)

Qwen3.5-9B-abliterated

by lukey03 · built on Qwen3.5 9B Base
#92
NewUncensored9BThinking
156Kdownloads112likes

Qwen3.5-9B-abliterated is a 9B model by lukey03 of Qwen3.5 9B Base, using abliteration. It can read images and writes out its reasoning before it answers. Languages noted on the card include Chinese, Japanese, Korean. The Q4_K_M file is 5.63 GB.

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

iOS-safe from 12 GB · Android-safe from 12 GB · 5.63 GB download · ~6–12 tok/s (estimate)

gemma-4-E4B-it-ultra-uncensored-heretic

by llmfan46 · built on Gemma 4 E4B
#21
Uncensored8B · 4.5B activeThinkingVision
Uncensored IQ 38.5
Reasoning16.4
Answers10/10
146Kdownloads329likes

gemma-4-E4B-it-ultra-uncensored-heretic is a 8B · 4.5B active model by llmfan46 of Gemma 4 E4B, using heretic, an automated abliteration tool. It can read images and writes out its reasoning before it answers. The Q4_K_M file is 5.34 GB.

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

iOS-safe from 12 GB · Android-safe from 8 GB · 5.34 GB download · ~6–12 tok/s (estimate)

Gemma-4-E2B-Uncensored-HauhauCS-Aggressive

by HauhauCS · built on Gemma 4 E2B
#73
Uncensored5.1B · 2.3B activeVision
114Kdownloads382likes

Gemma-4-E2B-Uncensored-HauhauCS-Aggressive is a 5.1B · 2.3B active model by HauhauCS of Gemma 4 E2B, using fine-tuning. It can look at a photo or screenshot and answer questions about it. The Q4_K_M file is 3.45 GB.

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

iOS-safe from 8 GB · Android-safe from 6 GB · 3.45 GB download · ~8–16 tok/s (estimate)

Qwen3.5-2B-Uncensored-HauhauCS-Aggressive

by HauhauCS · built on Qwen3.5 2B
#74
Uncensored2BVision
78Kdownloads185likes

Qwen3.5-2B-Uncensored-HauhauCS-Aggressive is a 2B model by HauhauCS of Qwen3.5 2B, using uncensoring. It can look at a photo or screenshot and answer questions about it. Languages noted on the card include Chinese. The Q4_K_M file is 1.27 GB.

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

iOS-safe from 4 GB · Android-safe from 4 GB · 1.27 GB download · ~18–30 tok/s (estimate)

Qwen3.8-9B-Distill-uncensored-heretic

by nurdich · built on Qwen3.5 9B
#88
NewUncensored9BThinkingVision
73Kdownloads37likes

Qwen3.8-9B-Distill-uncensored-heretic is a 9B model by nurdich of Qwen3.5 9B, using heretic, an automated abliteration tool. It can read images and writes out its reasoning before it answers. The Q4_K_M file is 5.78 GB.

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

iOS-safe from 12 GB · Android-safe from 8 GB · 5.78 GB download · ~6–12 tok/s (estimate)

Uncensored4BVision
60Kdownloads31likes

DeepSeek-V4-Pro-Qwen3.5-4B-UNCENSORED-heretic is a 4B model by xtools-at, using heretic, an automated abliteration tool. It can read images and writes out its reasoning before it answers. The Q4_K_M file is 2.78 GB.

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

iOS-safe from 6 GB · Android-safe from 4 GB · 2.78 GB download · ~12–22 tok/s (estimate)

Uncensored9BVision
41Kdownloads27likes

Qwen3.8-9B-heretic-uncensored is a 9B model by rohit267, using heretic, an automated abliteration tool. It can read images and writes out its reasoning before it answers. The Q4_K_M file is 5.63 GB.

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

iOS-safe from 12 GB · Android-safe from 8 GB · 5.63 GB download · ~6–12 tok/s (estimate)

Qwen3.5-9B-Ultra-Uncensored-Heretic

by llmfan46 · built on Qwen3.5 9B
#98
NewUncensored9BThinkingVision
31Kdownloads74likes

Qwen3.5-9B-Ultra-Uncensored-Heretic is a 9B model by llmfan46 of Qwen3.5 9B, using heretic, an automated abliteration tool. It can read images and writes out its reasoning before it answers. The Q4_K_M file is 5.53 GB.

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

iOS-safe from 8 GB · Android-safe from 8 GB · 5.53 GB download · ~6–12 tok/s (estimate)

Huihui-Qwen3.5-2B-abliterated

by huihui-ai · built on Qwen3.5 2B
#78
Uncensored2BVision
26Kdownloads10likes

Huihui-Qwen3.5-2B-abliterated is a 2B model by huihui-ai of Qwen3.5 2B, using abliteration. It can look at a photo or screenshot and answer questions about it. The Q4_K_M file is 1.27 GB.

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

iOS-safe from 4 GB · Android-safe from 4 GB · 1.27 GB download · ~18–30 tok/s (estimate)

Ministral-3-3B-Instruct-2512

by unsloth · built on Ministral 3 3B Base 2512
#24
Official3BVision
25Kdownloads44likes

An official 3B vision model based on Ministral 3 3B Base 2512, made to understand images as well as text. The Q4_K_M file is 2.15 GB.

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

iOS-safe from 4 GB · Android-safe from 4 GB · 2.15 GB download · ~12–22 tok/s (estimate)

NewUncensored8BVision
24Kdownloads111likes

Qwen3VL-8B-Uncensored-HauhauCS-Aggressive is a 8B model by HauhauCS, using uncensoring. It can look at a photo or screenshot and answer questions about it. Languages noted on the card include Chinese. The Q4_K_M file is 5.03 GB.

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

iOS-safe from 12 GB · Android-safe from 8 GB · 5.03 GB download · ~6–12 tok/s (estimate)

Qwen3.5-4B-SOMPOA-heresy-v2

by MuXodious · built on Qwen3.5 4B
#1
Uncensored4BThinkingVision
Uncensored IQ 44.7
Reasoning22.4
Answers10/10
19Kdownloads9likes

Qwen3.5-4B-SOMPOA-heresy-v2 is a 4B model by MuXodious of Qwen3.5 4B, using heretic, an automated abliteration tool. It can read images and writes out its reasoning before it answers. The Q4_K_M file is 2.71 GB.

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

iOS-safe from 4 GB · Android-safe from 4 GB · 2.71 GB download · ~12–22 tok/s (estimate)

Qwen3.5-9B-Claude-4.6-HighIQ-INSTRUCT-HERETIC-UNCENSORED

by DavidAU · built on Qwen3.5 9B heretic v2
#2030
NewUncensored9BThinkingVision
17Kdownloads60likes

Qwen3.5-9B-Claude-4.6-HighIQ-INSTRUCT-HERETIC-UNCENSORED is a 9B model by DavidAU of Qwen3.5 9B heretic v2, using heretic, an automated abliteration tool. It can read images and writes out its reasoning before it answers. The Q4_K_M file is 5.63 GB.

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

iOS-safe from 12 GB · Android-safe from 8 GB · 5.63 GB download · ~6–12 tok/s (estimate)

gemma-4-E2B-it-ultra-uncensored-heretic

by llmfan46 · built on Gemma 4 E2B
#81
Uncensored5.1B · 2.3B activeVision
15Kdownloads41likes

gemma-4-E2B-it-ultra-uncensored-heretic is a 5.1B · 2.3B active model by llmfan46 of Gemma 4 E2B, using heretic, an automated abliteration tool. It can look at a photo or screenshot and answer questions about it. The Q4_K_M file is 3.43 GB.

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

iOS-safe from 8 GB · Android-safe from 6 GB · 3.43 GB download · ~8–16 tok/s (estimate)

Gemma-3-4B-VL-it-Gemini-Pro-Heretic-Uncensored-Thinking

by DavidAU · built on gemma 3 4b heretic uncensored abliterated Extreme
#100
NewUncensored4BThinkingVision
10Kdownloads80likes

Gemma-3-4B-VL-it-Gemini-Pro-Heretic-Uncensored-Thinking is a 4B model by DavidAU of gemma 3 4b heretic uncensored abliterated Extreme, using heretic, an automated abliteration tool. It can read images and writes out its reasoning before it answers. The Q4_K_M file is 2.49 GB.

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

iOS-safe from 4 GB · Android-safe from 4 GB · 2.49 GB download · ~12–22 tok/s (estimate)

Qwen3.5-9B-heretic-v2

by trohrbaugh · built on Qwen3.5 9B Base
#2
Uncensored9BThinkingVision
Uncensored IQ 40.5
Reasoning24.2
Answers9.5/10
9.7Kdownloads3likes

Qwen3.5-9B-heretic-v2 is a 9B model by trohrbaugh of Qwen3.5 9B Base, using heretic, an automated abliteration tool. It can read images and writes out its reasoning before it answers. The Q4_K_M file is 5.63 GB.

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

iOS-safe from 12 GB · Android-safe from 8 GB · 5.63 GB download · ~6–12 tok/s (estimate)

Qwen3.5-9B-Claude-4.6-HighIQ-THINKING-HERETIC-UNCENSORED

by DavidAU · built on Qwen3.5 9B heretic v2
#26
Uncensored9BThinkingVision
Uncensored IQ 30.3
Reasoning22.9
Answers6.5/10
9.2Kdownloads25likes

Qwen3.5-9B-Claude-4.6-HighIQ-THINKING-HERETIC-UNCENSORED is a 9B model by DavidAU of Qwen3.5 9B heretic v2, using heretic, an automated abliteration tool. It can read images and writes out its reasoning before it answers. The Q4_K_M file is 5.63 GB.

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

iOS-safe from 12 GB · Android-safe from 8 GB · 5.63 GB download · ~6–12 tok/s (estimate)

Qwythos-9B-Claude-Mythos-5-1M-uncensored-heretic

by llmfan46 · built on Qwythos 9B Claude Mythos 5 1M
#99
NewUncensored9BThinkingVision
9Kdownloads94likes

Qwythos-9B-Claude-Mythos-5-1M-uncensored-heretic is a 9B model by llmfan46 of Qwythos 9B Claude Mythos 5 1M, using heretic, an automated abliteration tool. It can read images and writes out its reasoning before it answers. The Q4_K_M file is 6.22 GB.

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

iOS-safe from 12 GB · Android-safe from 12 GB · 6.22 GB download · ~6–12 tok/s (estimate)

NewUncensored9BThinkingVisionCode
8.7Kdownloads150likes

Qwen3.5-9B-Claude-4.6-OS-Auto-Variable-HERETIC-UNCENSORED-THINKING-MAX-NEOCODE-Imatrix is a 9B model by DavidAU of Qwen3.5 9B heretic v2, using heretic, an automated abliteration tool. It can read images and writes out its reasoning before it answers, and is set up for writing code. The Q4_K_M file is 6.84 GB.

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

iOS-safe from 12 GB · Android-safe from 12 GB · 6.84 GB download · ~6–12 tok/s (estimate)

Huihui-Qwen3.5-0.8B-abliterated

by huihui-ai · built on Qwen3.5 0.8B
#85
Uncensored0.8BVision
7.1Kdownloads10likes

Huihui-Qwen3.5-0.8B-abliterated is a 0.8B model by huihui-ai of Qwen3.5 0.8B, using abliteration. It can look at a photo or screenshot and answer questions about it. The Q4_K_M file is 0.53 GB.

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

iOS-safe from 4 GB · Android-safe from 4 GB · 0.53 GB download · ~25–45 tok/s (estimate)

gemma-4-E4B-it-heretic

by coder3101 · built on Gemma 4 E4B
#13
Uncensored8B · 4.5B activeThinkingVision
Uncensored IQ 37.7
Reasoning18.7
Answers9.8/10
3Kdownloads6likes

gemma-4-E4B-it-heretic is a 8B · 4.5B active model by coder3101 of Gemma 4 E4B, using heretic, an automated abliteration tool. It can read images and writes out its reasoning before it answers. The Q4_K_M file is 5.34 GB.

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

iOS-safe from 12 GB · Android-safe from 8 GB · 5.34 GB download · ~6–12 tok/s (estimate)

Crow-9B-HERETIC-4.6

by Crownelius · built on Qwen3.5 9B heretic v2
#18
Uncensored9BThinkingVision
Uncensored IQ 35.7
Reasoning19.5
Answers7.8/10
2.9Kdownloads287likes

Crow-9B-HERETIC-4.6 is a 9B model by Crownelius of Qwen3.5 9B heretic v2, using heretic, an automated abliteration tool. It can read images and writes out its reasoning before it answers. Languages noted on the card include Chinese, Russian, Spanish. The Q4_K_M file is 5.63 GB.

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

iOS-safe from 12 GB · Android-safe from 12 GB · 5.63 GB download · ~6–12 tok/s (estimate)

Qwen3.5-4B-ARA-heresy-v2

by MuXodious · built on Qwen3.5 4B
#3
Uncensored4BThinkingVision
Uncensored IQ 40.5
Reasoning22.6
Answers9/10
1.9Kdownloads1likes

Qwen3.5-4B-ARA-heresy-v2 is a 4B model by MuXodious of Qwen3.5 4B, using heretic, an automated abliteration tool. It can read images and writes out its reasoning before it answers. The Q4_K_M file is 2.71 GB.

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

iOS-safe from 4 GB · Android-safe from 4 GB · 2.71 GB download · ~12–22 tok/s (estimate)

Heretical-Qwen3.5-9B

by Kewk · built on Qwen3.5 9B Base
#64
Uncensored9BThinking
Uncensored IQ 30.7
Reasoning6.6
Answers7.8/10
829downloads1likes

Heretical-Qwen3.5-9B is a 9B model by Kewk of Qwen3.5 9B Base, using heretic, an automated abliteration tool. It can read images and writes out its reasoning before it answers. The Q4_K_M file is 5.63 GB.

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

iOS-safe from 12 GB · Android-safe from 8 GB · 5.63 GB download · ~6–12 tok/s (estimate)

qwen3-vl-8b-instruct-null-space-abliterated

by jwest33 · built on Qwen3 VL 8B
#32
Uncensored8BVision
Uncensored IQ 37.1
Reasoning14.5
Answers8.5/10
769downloads10likes

qwen3-vl-8b-instruct-null-space-abliterated is a 8B model by jwest33 of Qwen3 VL 8B, using community fine-tuning. It can look at a photo or screenshot and answer questions about it. The Q4_K_M file is 5.03 GB.

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

iOS-safe from 12 GB · Android-safe from 8 GB · 5.03 GB download · ~6–12 tok/s (estimate)

Qwen3.5-4B-PaperWitch-heresy-v2

by MuXodious · built on Qwen3.5 4B
#5
Uncensored4BThinkingVision
Uncensored IQ 41.7
Reasoning20.9
Answers9.5/10
754downloads2likes

Qwen3.5-4B-PaperWitch-heresy-v2 is a 4B model by MuXodious of Qwen3.5 4B, using heretic, an automated abliteration tool. It can read images and writes out its reasoning before it answers. The Q4_K_M file is 2.71 GB.

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

iOS-safe from 4 GB · Android-safe from 4 GB · 2.71 GB download · ~12–22 tok/s (estimate)

gemma-3-4b-it-null-space-abliterated

by jwest33 · built on Gemma 3 4B
#56
Uncensored4BVision
Uncensored IQ 31.2
Reasoning11.5
Answers8.2/10
732downloads1likes

gemma-3-4b-it-null-space-abliterated is a 4B model by jwest33 of Gemma 3 4B, using norm-preserving abliteration (MPOA). It can look at a photo or screenshot and answer questions about it. The Q4_K_M file is 2.49 GB.

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

iOS-safe from 6 GB · Android-safe from 4 GB · 2.49 GB download · ~12–22 tok/s (estimate)

Qwen-SEA-LION-v4-8B-VL-Magic_decensored

by MagicalAlchemist · built on Qwen SEA LION v4 8B VL
#10
Uncensored8B
Uncensored IQ 41.7
Reasoning15.8
Answers9.2/10
382downloads1likes

Qwen-SEA-LION-v4-8B-VL-Magic_decensored is a 8B model by MagicalAlchemist of Qwen SEA LION v4 8B VL, using heretic, an automated abliteration tool. It can look at a photo or screenshot and answer questions about it. Languages noted on the card include Vietnamese, Indonesian, Thai. The Q4_K_M file is 5.03 GB.

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

iOS-safe from 8 GB · Android-safe from 8 GB · 5.03 GB download · ~6–12 tok/s (estimate)

Qwen-SEA-LION-v4-4B-VL-Magic_decensored

by MagicalAlchemist · built on Qwen SEA LION v4 4B VL
#36
Uncensored4B
Uncensored IQ 36.3
Reasoning14.2
Answers8.5/10
265downloads1likes

Qwen-SEA-LION-v4-4B-VL-Magic_decensored is a 4B model by MagicalAlchemist of Qwen SEA LION v4 4B VL, using heretic, an automated abliteration tool. It can look at a photo or screenshot and answer questions about it. Languages noted on the card include Vietnamese, Indonesian, Thai. The Q4_K_M file is 2.5 GB.

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

iOS-safe from 4 GB · Android-safe from 4 GB · 2.5 GB download · ~12–22 tok/s (estimate)

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

Browse by memory

Browse by use

Quick answers

Does it cost anything?

No. The model files and PocketPal are free. There is no subscription, no API fee and no account. Some other apps, such as Layla, are paid. Licences differ between models (Apache 2.0, Gemma, Llama and a few non-commercial ones), so check the licence on a model's page before using it in a product.

Can a phone really run an AI model with no internet?

Yes. Every model here is a small language model that has been compressed (quantized) into a single GGUF file of roughly 0.5 to 6 GB. You download it once, and from then on your phone's own processor generates every answer. It keeps working in airplane mode, and nothing you type is sent anywhere.

Why run a model on your phone instead of a cloud chatbot?

It is free after the download, and it stays private. Your notes, ideas and questions never leave the device. There is no account, no subscription and no rate limit, and it still works on a plane, in the subway or anywhere without a signal.

How do I know if a model fits my phone?

Pick iOS or Android and your device memory in the bar at the top of the list. We treat only part of the RAM as usable, because the operating system and the app need the rest: about 57% on iOS (in the 55–60% range) and about 67% on Android (in the 65–70% range). A model counts as a fit when its Q4_K_M file, or a smaller Q3_K_M file, stays within that budget. These are rules of thumb, not a guarantee. Close other apps before you load a model, and if it will not load, try the smaller file or a smaller model.

How capable are phone-sized models?

Be realistic: they are noticeably weaker than large cloud or desktop models. The strongest uncensored model in this list scores 44.7 on the UGI Leaderboard, and the tiniest ones are best kept to quick lookups. They are useful for brainstorming, summarising and talking an idea through, but double-check anything that matters.

Read the full guide & FAQ →

Copied