interactive archive
Model extraction playground — version archive
Every snapshot below was recovered from git log on
which-models-extracted-playground.html in
playgroup_202602_docextract.
Filenames use the commit timestamp in UTC (YYYY-MM-DDTHHMMZ).
How to use: Click any row to open the playground here (embedded viewer). Use ← Playground archive to return to this list. Commit links open GitHub in a new tab. Open in a new tab: right-click a row or ⌘/Ctrl+click.
Latest snapshot
Current benchmark — 111 scored runs · 8 tabs · 2026-08-22T0830Z
| Snapshot (UTC) | Models | F1 runs | Providers | Avg F1 (active) | Tabs | Commit | Change |
|---|---|---|---|---|---|---|---|
| Loading… | |||||||
Historic snapshots
20 earlier versions recovered from git — click a row to open
| Snapshot (UTC) | Models | F1 runs | Providers | Avg F1 (active) | Tabs | Commit | Change |
|---|---|---|---|---|---|---|---|
| Loading versions… | |||||||
Milestones
- 2026-02-28 — first multi-model leaderboard (exact-match tabs)
- 2026-03-06 — time/cost estimation & Project Evolution tab
- 2026-03-07 — provider awareness, F1 scoring, Provider Analysis tab
- 2026-04-11 — V7 Go provider (85 models)
- 2026-07-29 — 107 scored runs, 34 Doubleword models
- 2026-08-22 — 111 scored runs, 8 tabs (latest; includes dw-deepseek-ocr-2)