Llama Server GUI
The featherweight, high-performance native desktop cockpit and flag configurator for llama-server and llama.cpp. Zero terminal friction. Pure silicon inference speed.
Everything You Need to Run Local Models
Eliminate confusing CLI arguments, batching flags, and CUDA errors.
Auto Hardware Tuning
Instantly detects NVIDIA CUDA GPUs, offloads all layers (-ngl 99), and enables Flash Attention (-fa on) automatically for peak tokens/sec.
Local Model Scanner
Automatically crawls your Downloads, Documents, and HuggingFace directories for .gguf model files. 1-click select and launch.
API Chat Playground
Test and converse with your local models right inside the GUI with real-time streaming tokens, time-to-first-token (TTFT), and latency performance metrics.
OpenAI Compatible API
Serves a drop-in local endpoint (http://127.0.0.1:8080/v1) compatible with Open WebUI, AnythingLLM, Continue.dev, Cursor, and custom code.
Ultra Lightweight (2.8 MB)
Built with Tauri v2 and Rust. Uses negligible RAM and disk space, completely ditching bloated Electron runtimes for pure native desktop speed.
Retro Noir & OLED Themes
Styled with amber terminal accents, OLED midnight blacks, and responsive shortcuts for power users (Ctrl+K, F1-F3).
Llama Server GUI v2.0.0
Windows 10 / 11 (x64) • NSIS Clean Installer • 2.8 MB • Open Source under Apache 2.0