Fast LLM inference.
Local, private, yours.

Native C++ and PyTorch inference. No cloud. No API keys. Run frontier models on your own hardware.

git clone https://github.com/LMGNU/llm.cpp

Get running in minutes.

Choose your preferred environment to get started.

# 1. Create a virtual environment
python -m venv .venv
.venv/Scripts/python.exe -m pip install --upgrade pip

# 2. Install dependencies
cd backend
../.venv/Scripts/python.exe -m pip install -r requirements.txt

# 3. Interactive chat (PyTorch inference)
python engine/inference.py --checkpoint "engine/best_model.pt"

# 4. Single generation
python engine/inference.py --checkpoint "engine/best_model.pt" --prompt "Hello" --max-new-tokens 100 --temperature 1.0
# 1. Build
g++ -std=c++17 -O2 -I. -Iinclude -o llm.exe main.cpp

# 2. Train from scratch
./llm.exe data/input.txt

# 3. Chat mode
./llm.exe data/input.txt --chat

# 4. Generation mode
./llm.exe data/input.txt --generate

Runs on your hardware.

Optimized for every major platform — from laptops to data centers.

Apple Silicon
NVIDIA RTX
NVIDIA H100 / A100
Intel iGPU
AMD GPU
CPU x86-64
CPU ARM
NVIDIA T4