LLM Hoster

JoblessAiDeveloper

About this app

Turn your Android phone into a powerful, private AI server — no cloud, no subscription, no data leaving your device.

LLM Hoster lets you download and run large language models (LLMs) directly on your phone using the llama.cpp inference engine, and exposes a fully compatible API over your local network. Use it with any client that supports the baseAPI/v1API — Open WebUI, Continue.dev, Chatbox, LM Studio, or your own apps.

━━━━━━━━━━━━━━━━━━━━

🔒 COMPLETELY PRIVATE
Your data never leaves your device. No internet required for inference. No API keys, no cloud costs, no telemetry. What happens on your phone stays on your phone.

📥 ONE-TAP MODEL DOWNLOADS
Browse and download GGUF models from Hugging Face with a single tap. Just paste a repo name like "author/model_name-GGUF" — the app automatically finds and downloads the best quantized version for mobile.

🌐 /v1-COMPATIBLE API
Expose a real REST API on your LAN:
• GET /v1/models
• POST /v1/chat/completions
• POST /v1/completions
Connect from your laptop, tablet, or any device on the same network.

💬 BUILT-IN CHAT
Chat with your loaded model directly inside the app. No external client needed for quick conversations.

⚡ PERFORMANCE TUNING
Fine-tune inference for your device:
• CPU core allocation (2–8 threads)
• KV cache quantization (q4_0, q8_0)
• Context window size (2K–16K tokens)
• Batch size control
• GPU layer offloading
• Flash Attention support

🔧 RUNS IN THE BACKGROUND
A foreground service keeps the server alive even when you switch apps. Stop it anytime from the notification shade.

━━━━━━━━━━━━━━━━━━━━

SUPPORTED MODELS
Works with any GGUF-format model. Recommended starting points:
• Llama 3.2 (1B / 3B)
• Phi-3 Mini
• Gemma 2 (2B)
• Qwen 2.5 (1.5B / 3B)
• Mistral 7B (Q4 quantized)

Smaller quantized models (Q4_K_M, Q4_0) run best on phones. Models are stored locally and persist across sessions.

━━━━━━━━━━━━━━━━━━━━

USE CASES
• Private AI assistant with zero cloud dependency
• Development & testing of LLM-powered apps
• Offline AI in areas with limited connectivity
• Learning and experimenting with open-source models
• Serving AI to multiple devices on your home network

━━━━━━━━━━━━━━━━━━━━

No ads. No tracking. No accounts. Just AI on your phone.

Download LLM Hoster and start hosting your own private AI server today.

What’s new

• Advanced Settings: New controls for GPU offloading, thread batches, and speculative decoding.
• Stability improvements and general UI cleanup.