Skip to content
fedi.software

fedi-index: Language: English

Best models for this language by Arena votes in it. Provisional: one source family.

Live data for 2026-10 (recalculated daily) · Methodology v1 · Methodology · History · JSON

All Open weights Closed
# Tier Model fedi-index Percentile Value Sources Score decomposition
6A1,476 ±1598 +93◆ A›
15A1,458 ±695 +44 A›
17A1,458 ±794 +60 A›
23A1,454 ±592 +80 A›
25A1,452 ±1291 +98◆ A›
26A1,452 ±591 +36 A›
28A1,451 ±1090 +92◆ A›
29A1,451 ±590 +45 A›
30A1,449 ±789 +105◆ A›
38A1,436 ±686 +44 A›
39A1,436 ±586 +33 A›
42A1,434 ±685 +23 A›
48A1,430 ±983 +33 A›
50A1,430 ±482 +36 A›
53A1,429 ±981 +20 A›
62B1,426 ±1278 +59 A›
63B1,425 ±977 +68 A›
64B1,425 ±577 +26 A›
67B1,424 ±576 +75 A›
71B1,422 ±874 +46 A›
73B1,420 ±474 +27 A›
74B1,419 ±673 +27 A›
75B1,418 ±573 +50 A›
77B1,418 ±1272 +54 A›
79B1,417 ±571 +71 A›
81B1,416 ±971 +46 A›
88B1,413 ±668 +18 A›
91B1,409 ±867 -19 A›
92B1,408 ±966 +19 A›
94B1,406 ±566 +55◆ A›
95B1,405 ±665 +14 A›
97B1,404 ±965 +27 A›
98B1,404 ±1564 +31 A›
100B1,402 ±464 +32 A›
101B1,401 ±663 +14 A›
103B1,398 ±662 +32 A›
104B1,397 ±662 +38 A›
105B1,395 ±1062 -17 A›
108B1,394 ±561 +50 A›
117B1,383 ±657 +11 A›
121B1,379 ±756 +18 A›
122B1,376 ±1056 +0 A›
124B1,374 ±655 -2 A›
127B1,369 ±854 -10 A›
128B1,368 ±1653 -12 A›
129B1,367 ±653 +25 A›
141C1,357 ±548 +9 A›
144C1,355 ±947 -22 A›
146C1,354 ±847 +6 A›
148C1,352 ±646 -2 A›
149C1,350 ±846 +1 A›
155C1,340 ±643 -36 A›
156C1,339 ±1143 -32 A›
157C1,338 ±1243 -60 A›
159C1,336 ±642 -22 A›
161C1,334 ±741 -14 A›
162C1,333 ±1541 -40 A›
163C1,333 ±640 -30 A›
170C1,320 ±1438 — A›
171C1,312 ±638 -29 A›
172C1,311 ±1637 -14 A›
177C1,298 ±735 — A›
180C1,292 ±1034 -95 A›
181C1,291 ±1434 -28 A›
184C1,280 ±933 -26 A›
209D1,191 ±1224 -133 A›
214D1,166 ±622 -236 A›

…

AArena (human votes) ◆ Pareto frontier: nothing is both better and cheaper

fedi-index is an Elo-equivalent on a fixed scale: comparable between months within one methodology version. Tiers S–D: by percentile in the slice, and no better than the confidence interval allows (cut-offs in the methodology). This slice has one source family, so all its scores are provisional. Methodology