Loading…
CompareOpen in Google Play
This app isn't in any of our charts yet.
Version 1.2.0 • NEW: Qwen3 8B (Q3_K_M) model support via llama.cpp engine • NEW: GGUF model format support — run any llama.cpp-compatible model • NEW: In-app chat — test your LLM directly from the Server tab • NEW: Syntax-highlighted API docs with copy button (cURL, Python, JS) • NEW: Anthropic /v1/messages endpoint support • IMPROVED: Multi-engine architecture (LiteRT-LM + llama.cpp) • IMPROVED: Model catalog with one-tap download • FIXED: Model switching no longer requires server restart
Run Gemma & Qwen3 AI models locally with an OpenAI-compatible API. GGUF support.
Transform your Android device into a powerful local AI server with LLM Server. Download and host Large Language Models (LLMs) securely on your device and serve them via a local REST API compatible with standard OpenAI client libraries. Whether you're an AI enthusiast, developer, or privacy advocate, enjoy advanced AI completions without relying on cloud services. Your data never leaves your device! Key Features: • 100% Offline AI: Run inference entirely on-device — your prompts, data, and responses remain strictly private. • OpenAI API Compatible: Plug into existing LLM pipelines, chat apps, and scripts. Supports /v1/chat/completions, /v1/completions, and Anthropic /v1/messages endpoints. • Multi-Engine Support: Run Google Gemma models via LiteRT-LM with GPU acceleration, or GGUF models (like Qwen3 8B) via llama.cpp — choose the best model for your device. • Built-in Model Catalog: One-tap download of Gemma 4 E4B (4B, fast) or Qwen3 8B (Q3_K_M, smarter). Or load any compatible model file. • In-App Chat: Test your local LLM directly from the Server tab — no external tools needed. • API Documentation: Syntax-highlighted code examples for cURL, Python, and JavaScript with one-tap copy. • Customizable Parameters: Adjust Max Tokens, Top K, and Temperature from the dashboard. • API Key Rotation: Secure your local server with rotating access tokens. • Premium Subscription: Unlock unlimited server uptime. Supported Models: • Gemma 4 E4B — Google's compact 4B model, LiteRT-LM with GPU acceleration • Qwen3 8B (Q3_K_M) — Alibaba's powerful 8B model via GGUF/llama.cpp (~4.1 GB) • Any GGUF-format model compatible with llama.cpp Getting Started: 1. Select a model from the catalog and download it. 2. Start the server. 3. Use the built-in chat or copy your API URL and Key to your favorite client! Note: Performance depends on your device's RAM and processor. System Requirements: • 💾 Minimum 8 GB RAM (12 GB+ recommended for 8B models) • ⚡ Snapdragon 8 Gen 2 or above recommended Performance depends on your device's processor and available memory. Devices with less RAM may experience slow inference or fail to load larger models.