About Gautam Ramachandra
I’m building Banach, a free, open-source tool that runs very large AI models (Mixture-of-Experts models like DeepSeek and Qwen) on regular laptops and desktops, even when the model is bigger than your RAM.
Before you download anything, it tells you how fast the model will run on your machine. While it runs, it gives the same answers as standard llama.cpp, and it measures any shortcut it takes so you know what you’re trading off.
Your support pays for:
test hardware (GPUs, a big-RAM machine for the 284B-model benchmarks)
CI time and storage for model downloads
time to keep it working on Linux, macOS, Windows, NVIDIA, AMD and Apple GPUs
Every coffee keeps this free and open for everyone. Thank you!
Before you download anything, it tells you how fast the model will run on your machine. While it runs, it gives the same answers as standard llama.cpp, and it measures any shortcut it takes so you know what you’re trading off.
Your support pays for:
test hardware (GPUs, a big-RAM machine for the 284B-model benchmarks)
CI time and storage for model downloads
time to keep it working on Linux, macOS, Windows, NVIDIA, AMD and Apple GPUs
Every coffee keeps this free and open for everyone. Thank you!
Recent supporters