eGPU

16GB VRAM Runs 200K Context: Local LLMs Cross a Real Threshold

A Reddit user ran a 27B-parameter Qwen model at 200K-token context on a laptop with an eGPU. Mid-size AI just got much closer to a regular desktop.