Written by Jakub Rusinowski · Last updated July 10, 2026
Your smartphone is more powerful than you think. Modern flagship phones run 1B–3B parameter LLMs entirely on-device, with no internet required, no API keys, and full privacy. This guide covers everyth
Your smartphone is more powerful than you think. Modern flagship phones run 1B–3B parameter LLMs entirely on-device, with no internet required, no API keys, and full privacy. This guide covers everything you need to get started.
| Phone RAM | Models You Can Run | Practical Speed |
|---|---|---|
| 6GB | SmolLM2 1.7B, Phi-3 Mini (limited) | 5–10 tokens/sec |
| 8GB | Llama 3.2 1B, SmolLM2 1.7B, Gemma 3 1B | 8–15 tokens/sec |
| 12GB | Llama 3.2 3B, Phi-3 Mini 3.8B, Gemma 2 2B | 10–20 tokens/sec |
| 16GB | Llama 3.2 3B (Q8), Phi-3.5 Mini, Qwen2.5 3B | 12–25 tokens/sec |
> Realistic expectation: Phones run small (1B–4B) models at chat speeds. They won't match a desktop with an RTX 4090, but they're genuinely useful for on-the-go tasks — and work 100% offline.
| App | Free? | Best For | Models Supported |
|---|---|---|---|
| PocketPal AI | Free | General chat, model variety | GGUF (Llama, Phi, Gemma, Qwen) |
| LLM Farm | Free | Privacy-focused, local-first | GGUF format |
| Enchanted | Free | Clean UI, Ollama remote or local | Llama, Mistral, Gemma |
| AI Chat — Local Models | Free/IAP | Beginner-friendly | Curated model list |
| App | Free? | Best For | Models Supported |
|---|---|---|---|
| PocketPal AI | Free | Cross-platform, best model library | GGUF (all popular models) |
| MLC Chat | Free | GPU-accelerated (Vulkan), fastest | MLC-converted models |
| Ollama (Android beta) | Free | Power users, API access | All Ollama models |
| Jan Mobile | Free | Open-source, OpenAI-compatible API | GGUF format |
Recommendation: Start with PocketPal AI on both platforms — it's free, cross-platform, actively maintained, and has the largest model compatibility.
| Model | Size | Min RAM | Best For | Download |
|---|---|---|---|---|
| Llama 3.2 1B Instruct | ~0.8GB | 4GB | Fast chat, simple tasks | PocketPal built-in |
| Llama 3.2 3B Instruct | ~2GB | 8GB | Best balance on phones | PocketPal built-in |
| Gemma 3 1B Instruct | ~0.7GB | 4GB | Multilingual, Google quality | PocketPal built-in |
| Phi-3 Mini 3.8B | ~2.4GB | 8GB | Reasoning, coding tasks | PocketPal / MLC |
| SmolLM2 1.7B Instruct | ~1GB | 4GB | Ultra-fast, efficient | PocketPal built-in |
| Qwen2.5 3B Instruct | ~2GB | 8GB | Multilingual, code | PocketPal / MLC |
PocketPal AI is the easiest and most capable local AI app for both platforms.
Step 1: Install
Step 2: Download a model 1. Open PocketPal → tap Models tab 2. You'll see a curated list of phone-optimized models 3. For first-time users: tap Llama 3.2 3B Instruct Q4 → Download 4. Wait for download (typically 2–4GB over WiFi)
Step 3: Start chatting 1. Tap Chat → select your downloaded model 2. Wait ~5 seconds for model to load into RAM 3. Type your message and send
Step 4: (Optional) Load custom GGUF models 1. Download any .gguf file from Hugging Face to your Files app (iOS) or Downloads (Android) 2. In PocketPal → Models → Add Custom Model → pick the file 3. Set context size (512–2048 recommended for phones)
LLM Farm is a privacy-focused, open-source app with fine-grained control.
Install: GitHub — LLM Farm or App Store search "LLM Farm"
Setup: 1. Open LLM Farm → tap + to add a model 2. Choose Download from HuggingFace or Import from Files 3. Recommended model string for beginners: `` https://huggingface.co/bartowski/Llama-3.2-3B-Instruct-GGUF/resolve/main/Llama-3.2-3B-Instruct-Q4_K_M.gguf `` 4. Set Context Size: 2048 (safe for 8GB phones), 4096 (12GB+ phones) 5. Set Threads: 4–6 (match your CPU core count) 6. Tap Save → select model → Chat
MLC Chat uses Vulkan GPU acceleration making it 2–3x faster than CPU-only apps on Android.
Install: GitHub — MLC-LLM Android (APK from GitHub releases)
Setup: 1. Install the APK (enable 'Install unknown apps' in Android settings) 2. Open MLC Chat → tap + to browse built-in models 3. Select Llama-3.2-3B-Instruct-q4f16_1 (best for most Android phones) 4. Download model (~2GB) → tap to load → Chat
Why MLC is faster: MLC pre-compiles models with TVM into GPU-native code, using your Adreno or Mali GPU via Vulkan rather than CPU. On a Snapdragon 8 Elite, expect 15–25 tokens/sec vs 8–12 on CPU-only apps.
iPhones have a dedicated Neural Engine with up to 38 TOPS (A18 Pro). Apps like PocketPal and LLM Farm automatically use it via Apple's Core ML framework:
Snapdragon 8 Gen 3 / 8 Elite phones have excellent AI acceleration via Hexagon NPU:
| Phone | App | Model | Speed | Battery Drain |
|---|---|---|---|---|
| iPhone 16 Pro | PocketPal | Llama 3.2 3B Q4 | ~15 t/s | ~6%/hour |
| iPhone 15 Pro | LLM Farm | Phi-3 Mini Q4 | ~12 t/s | ~8%/hour |
| Samsung S25 Ultra | MLC Chat | Llama 3.2 3B | ~22 t/s | ~10%/hour |
| Pixel 9 Pro XL | PocketPal | Llama 3.2 3B Q4 | ~14 t/s | ~9%/hour |
| OnePlus 13 | MLC Chat | Phi-3 Mini | ~20 t/s | ~10%/hour |
Understanding what phones can't do helps set the right expectations:
→ Check your phone's compatibility in the Hardware Analyzer | → Browse tiny models for phones