Stop wrestling with command-line flags. Start forging intelligence.
LlamaForge is a professional-grade Graphical User Interface (GUI) for llama.cpp. It transforms the complex process of running local Large Language Models (LLMs) into a streamlined, point-and-click experience without sacrificing the power-user controls you need.
- 🧠 Smart Runtime Detection: LlamaForge automatically scans your system for compatible DLLs (CUDA, ROCm, Vulkan) and only enables valid backends. No more crashes due to missing drivers.
- ⚡ Instant Backend Forcing: Want to force Vulkan over CUDA? CPU over ROCm? LlamaForge handles the complex environment variable overrides (
CUDA_VISIBLE_DEVICES,HIP_VISIBLE_DEVICES) instantly. - 📡 Real-Time Telemetry: View color-coded, streaming server logs directly in the UI. Track tokens per second, context loading, and errors as they happen.
- 📂 Model Arsenal: Recursively scans your directories for
.gguffiles. Select, load, or delete models from a clean dropdown menu. - 🛠️ Command Forge: See the exact command being generated. Edit it manually before execution for total control.
- ⬇️ Built-in Downloader: integrated access to HuggingFace for grabbing new models.
No Python. No Dependencies. It just works.
- Download the latest
LlamaForge_Beta.zipfrom the Releases page. - Extract the folder.
- Run
LlamaForge.exe. - Point it to your
llama-server.exe(part of llama.cpp) and your Model folder. - Forge.
- Select Runtime: Choose between Auto, CUDA (NVIDIA), ROCm (AMD), Vulkan, or CPU.
- Pick a Model: Select a
.gguffile from your scanned list. - Tune Parameters: Adjust Context Size, GPU Layers, and Threads with visual sliders.
- Launch: Click "Start Server". The dashboard lights up with your API endpoint and live logs.
If LlamaForge has saved you time or helped you run your local AI setup, consider supporting the development!
Built for the community, by the community.