Archived benchmark history (to 2026-09-28, retch v0.20.6). The current dashboard is one level up. Read this data with care:

• fastfetch series named fastfetch, fastfetch -c none, fastfetch -c all are not comparable with retch. They ran fastfetch's built-in default, no config, every module, or the developer's own config, not the fields retch shows. From v0.19.0 the series fastfetch (short|default|long|full) use matched per-mode configs and are comparable.
• retch series are retch over time, but the modes changed: v0.19.0 moved cpu-usage from the default mode to --long, so part of that drop is a redefinition, not a speed-up.
• CI absolute times vary a lot between runners (macOS retch default ranged 334-2596 ms); compare ratios within a run, not absolute values across runs.
• Local points carry no AC/battery state, which on a laptop moved --short by ~0.3 ms.