Running GPT-2 vs Qwen3 locally reveals 11x speed gap, not 297x

A developer ran GPT-2 XL (2019, 1.5B params) and Qwen3-0.6B (2025) on the same laptop using AMD's Lemonade server. Initial speed tests showed a 297x gap, but a reproducible benchmark corrected it to 11x. The project also led to the developer's first open-source contribution after hitting a documentation wall.