Apps & Tools
smolmaxxing
Run large language models on everyday low-memory hardware like TV boxes and old iPads.
Is this your creation?
PlayableOpen source

Description
An engineering project and llama.cpp patch that enables running models up to 35B parameters on devices with limited RAM. It utilizes custom kernels for 1-bit and ternary quantization, drops unused vocabulary rows, and implements flash-based paging for Mixture-of-Experts (MoE) layers.
Descriptions, tags, and model credits may be AI-generated or inferred from public sources and can be incomplete or wrong. Learn more.