swarchanalysis

architecture-extraction run comparison

Each repo was analyzed by up to four runs — Local (deepseek-v4-flash), Hybrid-flash (flash + v4-pro), OpenAI (OpenAI models), and Claude Code (opus-4.8). Repos are grouped by difficulty band; the badge is the course grade (0–10) awarded by the professor (Softeng projects are ungraded). Repos with two or more runs open a side-by-side comparison.

Loading…