Rocket - Think it. Type it. Launch it.
Mini toolsAirLLM
By 0xSojalSec
AirLLM reduces peak GPU memory by executing models layer-by-layer (keeping most weights off GPU until needed), enabling constrained-hardware inference without requiring model compression as the core premise of the project.
New
Text
Gen 7
Overview
AirLLM is an inference system that enables running very large LLMs on low-VRAM GPUs by loading model layers from disk in a memory-optimized execution pipeline.
Tools using AirLLM
No tools found for this model yet.
Last updated: March 12, 2026
