Start with your RAM
Every model page lists the minimum and recommended memory. Tiny models from 1.5–3 GB run on most phones; the 4B-class models want 6–8 GB; the largest need a flagship with 13 GB or more.
← Back to Offline AI Chat
EVERY MODEL, ON YOUR PHONE.
The model catalog
43 AI models you can download and run entirely on your phone — 17 free, 26 in Offline Plus — from 16 makers including Google, Hugging Face, IBM, Liquid AI and more. Search by name, skill or how much memory your phone has, then open a model page for download size and requirements.
Your phone's RAM
Free
SmolLM2 135M
Hugging Face · Simple questions
0.11 GB · from 1.5 GB RAM
Free
Qwen 3 0.6B
Qwen · Everyday tasks
0.64 GB · from 2.5 GB RAM
Free
Gemma 3 1B
Google · Writing
0.81 GB · from 3 GB RAM
Free
Qwen Coder 1.5B
Qwen · Coding
1.1 GB · from 3.5 GB RAM
Free
Qwen 3 1.7B
Qwen · Conversation
1.8 GB · from 4.5 GB RAM
Offline Plus
Qwen 3 4B
Qwen · Reasoning
2.5 GB · from 6 GB RAM
Free
LFM 2.5 230M
Liquid AI · Conversation
0.15 GB · from 3 GB RAM
Free
LFM 2.5 350M
Liquid AI · Conversation
0.23 GB · from 3 GB RAM
Free
LFM 2.5 Instruct 1.2B
Liquid AI · Conversation
0.73 GB · from 3 GB RAM
Offline Plus
LFM 2.5 Thinking 1.2B
Liquid AI · Reasoning
0.73 GB · from 3 GB RAM
Offline Plus
LFM 2.5 2.6B
Liquid AI · Conversation
1.7 GB · from 4.5 GB RAM
Free
Qwen 3.5 0.8B
Qwen · Conversation
0.53 GB · from 3 GB RAM
Free
Qwen 3.5 2B
Qwen · Conversation
1.3 GB · from 4.5 GB RAM
Offline Plus
Qwen 3.5 4B
Qwen · Conversation
2.7 GB · from 6 GB RAM
Offline Plus
Gemma 4 QAT E2B
Google · Conversation
3.3 GB · from 6 GB RAM
Offline Plus
Gemma 4 QAT E4B
Google · Conversation
5.2 GB · from 10 GB RAM
Offline Plus
Ministral 3 Instruct 3B
Mistral AI · Conversation
2.1 GB · from 4.5 GB RAM
Offline Plus
Ministral 3 Reasoning 3B
Mistral AI · Reasoning
2.1 GB · from 6 GB RAM
Free
SmolLM3 3B
Hugging Face · Conversation
1.9 GB · from 4.5 GB RAM
Free
Granite 4 Nano h-1b
IBM · Conversation
0.90 GB · from 3 GB RAM
Free
Granite 4 Micro 3B
IBM · Conversation
2.1 GB · from 4.5 GB RAM
Offline Plus
Phi-4 Mini 3.8B
Microsoft · Conversation
2.5 GB · from 6 GB RAM
Offline Plus
Qwen 3 Instruct 4B
Qwen · Conversation
2.5 GB · from 6 GB RAM
Offline Plus
LFM 2.5 MoE 8B-A1B
Liquid AI · Conversation
5.2 GB · from 10 GB RAM
Offline Plus
Qwen 3.5 9B
Qwen · Conversation
5.7 GB · from 13 GB RAM
Offline Plus
Qwen 3.5 Abliterated 0.8B
huihui-ai · Experimental
0.53 GB · from 3 GB RAM
Offline Plus
Qwen 3.5 Abliterated 2B
huihui-ai · Experimental
1.3 GB · from 4.5 GB RAM
Offline Plus
Qwen 3.5 Abliterated 4B
huihui-ai · Experimental
2.7 GB · from 6 GB RAM
Offline Plus
Qwen 3 Abliterated 1.7B
mlabonne · Experimental
1.1 GB · from 4.5 GB RAM
Offline Plus
Gemma 4 Abliterated E2B
huihui-ai · Experimental
3.4 GB · from 6 GB RAM
Offline Plus
LFM 2.5 Heretic 2.6B
noctrex · Experimental
1.7 GB · from 4.5 GB RAM
Free
MiniCPM 5 1B
OpenBMB · Conversation
0.69 GB · from 3 GB RAM
Free
MiniCPM 5 2B
OpenBMB · Conversation
1.6 GB · from 4.5 GB RAM
Free
Llama 3.2 1B
Meta · Conversation
0.81 GB · from 3 GB RAM
Free
Gemma 3 270M
Google · Simple questions
0.25 GB · from 2 GB RAM
Offline Plus
Llama 3.2 3B
Meta · Conversation
2.0 GB · from 4.5 GB RAM
Offline Plus
Llama 3.1 8B
Meta · Conversation
4.9 GB · from 10 GB RAM
Offline Plus
Nanbeige 4.2 3B
Nanbeige · Reasoning
2.7 GB · from 6 GB RAM
Offline Plus
Hy-MT2 Translator 1.8B
Tencent · Translation
1.1 GB · from 3.5 GB RAM
Offline Plus
Nemotron 3 Nano 4B
NVIDIA · Conversation
2.8 GB · from 6 GB RAM
Offline Plus
NeoHorse 1 4B
TokenRhythm · Coding
2.7 GB · from 6 GB RAM
Offline Plus
NeoHorse 1 Abliterated 4B
huihui-ai · Experimental
2.8 GB · from 6 GB RAM
Offline Plus
Llama 3.1 Abliterated 8B
mlabonne · Experimental
4.9 GB · from 10 GB RAM
No model matches that search. Try a smaller word, or clear the filters.
How to choose a model
Watch the download size
Quantized GGUF files run from about 0.1 GB to roughly 6 GB. Downloads continue in the background and can be paused, and each file is checked against its exact size and SHA-256 hash before it becomes available.
Then disconnect
Once a model is on your phone, chat works in airplane mode. Prompts and replies never leave the device, whatever model you pick.