AI-Roundtable Leaderboard
Ranking run · 2026-05-29

Ranking run 2026-05-29

Claude Opus 4.7 leads with 79/100 across all domains. Tax law: Claude Opus 4.7 (71) · Medicine: Claude Opus 4.7 (84) · Law: Claude Opus 4.7 (82) · Business law: GPT-5 (80).

Status
official run (feed, archive, homepage)
Run
run-20260529T043500-b4ffe0
Started
2026-05-29 04:35 UTC
Items
100 (Tax law 15, Medicine 20, Law 35, Business law 30)
Models
6
Judges
GPT-5, Claude Opus 4.8, Claude Opus 4.7, Mistral Large 2
Question bank
2026-05-17-v1
Code version
0.2.3
Pipeline cost
72.88 USD
Judge agreement
69 %

Overall across domains

#Modeltop10095 % CIΔ prev.failuressec median/p90hallu bandper 1000 tok.Δ own family
1Claude Opus 4.7019 / 317977–82—571,54—
2GPT-5034 / 597571–79—670,67—
3Claude Opus 4.8018 / 287471–77—720,91—
4Gemini 2.5 Pro021 / 265551–59—803,45—
5Mistral Large 209 / 655147–57—474,62—
6Grok 4.306 / 104743–51—764,80—

Δ prev. = difference to the previous published run; “—” = model was not in it. Hallu band: 100 = none, 0 = ≥ 4 false claims per answer. Per 1000 tok. = flagged claims per 1000 output tokens (length-normalised). Δ own family = score with all judges minus score with foreign judges only, in points; positive = own family scores more generously; “—” = not recorded before code_version 0.2.8 or no foreign judge (methodology §4, §6.5).

Tax law · 15 Items

Ranks 1–3 statistically tied (95 % CI overlaps rank 1)

In Tax law, Claude Opus 4.7 (top score) contradicts a stored key fact on average 0,5 times per answer and adds avg 2,0 claims the judge panel flags as false or unsupported; 100 % of answers carry at least one flag. The lowest value is Claude Opus 4.8 at avg 1,4 per answer. Relative to answer length: 1,95 per 1000 output tokens, lowest GPT-5 at 0,83.

#Modeltop10095 % CIΔ prev.contradicts key factextra false/unsupp.avg totalanswers affectedper 1000 tok.IRR
1Claude Opus 4.77166–77—0,52,02,4100 %1,9578 %
2Claude Opus 4.86963–74—0,60,91,4100 %1,0279 %
3GPT-56352–73—0,91,32,2100 %0,8369 %
4Gemini 2.5 Pro3828–48—1,20,61,880 %8,0360 %
5Mistral Large 22819–39—1,91,63,493 %7,5550 %
6Grok 4.32718–37—1,51,02,4100 %10,9860 %

Medicine · 20 Items

Ranks 1–3 statistically tied (95 % CI overlaps rank 1)

In Medicine, Claude Opus 4.7 (top score) contradicts a stored key fact on average 0,5 times per answer and adds avg 0,8 claims the judge panel flags as false or unsupported; 100 % of answers carry at least one flag. The lowest value is Gemini 2.5 Pro at avg 0,4 per answer. Relative to answer length: 1,15 per 1000 output tokens, lowest Claude Opus 4.8 at 0,68.

#Modeltop10095 % CIΔ prev.contradicts key factextra false/unsupp.avg totalanswers affectedper 1000 tok.IRR
1Claude Opus 4.78477–90—0,50,81,1100 %1,1581 %
2GPT-58172–89—0,60,51,080 %0,7578 %
3Claude Opus 4.87972–84—0,40,40,780 %0,6872 %
4Mistral Large 26760–74—0,40,91,295 %2,4770 %
5Gemini 2.5 Pro6152–67—0,10,30,460 %1,7173 %
6Grok 4.35547–64—0,30,20,460 %2,1875 %

Law · 35 Items

Ranks 1–3 statistically tied (95 % CI overlaps rank 1)

In Law, Claude Opus 4.7 (top score) contradicts a stored key fact on average 0,2 times per answer and adds avg 0,8 claims the judge panel flags as false or unsupported; 86 % of answers carry at least one flag. The lowest value is Gemini 2.5 Pro at avg 0,2 per answer. Relative to answer length: 1,22 per 1000 output tokens, lowest GPT-5 at 0,58.

#Modeltop10095 % CIΔ prev.contradicts key factextra false/unsupp.avg totalanswers affectedper 1000 tok.IRR
1Claude Opus 4.78277–86—0,20,81,086 %1,2282 %
2Claude Opus 4.87974–84—0,20,40,666 %0,6176 %
3GPT-57366–80—0,40,61,080 %0,5870 %
4Gemini 2.5 Pro6258–67—0,10,10,234 %1,2168 %
5Grok 4.34842–54—0,30,20,434 %3,6067 %
6Mistral Large 24638–54—1,30,71,977 %6,5364 %

Business law · 30 Items

Ranks 1–3 statistically tied (95 % CI overlaps rank 1)

In Business law, GPT-5 (top score) contradicts a stored key fact on average 0,4 times per answer and adds avg 1,0 claims the judge panel flags as false or unsupported; 97 % of answers carry at least one flag. The lowest value is Grok 4.3 at avg 1,1 per answer. Relative to answer length: 0,63 per 1000 output tokens.

#Modeltop10095 % CIΔ prev.contradicts key factextra false/unsupp.avg totalanswers affectedper 1000 tok.IRR
1GPT-58073–86—0,41,01,497 %0,6375 %
2Claude Opus 4.77772–81—0,32,22,5100 %1,7668 %
3Claude Opus 4.86863–74—0,51,41,997 %1,1862 %
4Mistral Large 25952–65—0,71,62,397 %3,6258 %
5Gemini 2.5 Pro5245–58—0,70,61,390 %4,2961 %
6Grok 4.35043–57—0,60,61,187 %4,1254 %

Raw data of this run

raw.jsonl and progress.jsonl are the public streams with response texts stripped (contamination protection, methodology §7).