Ranking run 2026-05-12
Claude Opus 4.7 leads with 77/100 across all domains. Tax law: Claude Opus 4.7 (68) · Medicine: GPT-5 (86) · Law: Claude Opus 4.7 (78) · Business law: Claude Opus 4.7 (75).
Overall across domains
| # | Model | top100 | 95 % CI | Δ prev. | failures | sec median/p90 | hallu band | per 1000 tok. | Δ own family |
|---|---|---|---|---|---|---|---|---|---|
| 1 | Claude Opus 4.7 | 0 | 17 / 26 | 77 | 71–81 | -1 | 54 | 1,45 | — |
| 2 | GPT-5 | 0 | 46 / 71 | 69 | 63–76 | -3 | 63 | 0,67 | — |
| 3 | Gemini 2.5 Pro | 0 | 22 / 28 | 47 | 40–54 | -2 | 72 | 4,20 | — |
| 3 | Mistral Large 2 | 0 | 16 / 372 | 47 | 40–55 | -2 | 45 | 3,99 | — |
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 · 8 Items
Ranks 1–2 statistically tied (95 % CI overlaps rank 1)
In Tax law, Claude Opus 4.7 (top score) contradicts a stored key fact on average 0,6 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 GPT-5 at avg 2,1 per answer. Relative to answer length: 1,95 per 1000 output tokens, lowest GPT-5 at 0,75.
| # | Model | top100 | 95 % CI | Δ prev. | contradicts key fact | extra false/unsupp. | avg total | answers affected | per 1000 tok. | IRR |
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Claude Opus 4.7 | 68 | 60–76 | +1 | 0,6 | 2,0 | 2,6 | 100 % | 1,95 | 80 % |
| 2 | GPT-5 | 56 | 45–67 | +6 | 1,0 | 1,1 | 2,1 | 100 % | 0,75 | 73 % |
| 3 | Gemini 2.5 Pro | 25 | 14–35 | -6 | 1,5 | 0,7 | 2,2 | 100 % | 9,92 | 64 % |
| 4 | Mistral Large 2 | 17 | 10–25 | 0 | 2,0 | 1,3 | 3,3 | 100 % | 7,15 | 61 % |
Medicine · 9 Items
Ranks 1–2 statistically tied (95 % CI overlaps rank 1)
In Medicine, GPT-5 (top score) contradicts a stored key fact on average 0,3 times per answer and adds avg 0,6 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 0,3 per answer. Relative to answer length: 0,67 per 1000 output tokens.
| # | Model | top100 | 95 % CI | Δ prev. | contradicts key fact | extra false/unsupp. | avg total | answers affected | per 1000 tok. | IRR |
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | GPT-5 | 86 | 81–91 | +4 | 0,3 | 0,6 | 0,9 | 100 % | 0,67 | 80 % |
| 2 | Claude Opus 4.7 | 85 | 80–90 | +3 | 0,2 | 0,6 | 0,8 | 89 % | 0,79 | 79 % |
| 3 | Mistral Large 2 | 69 | 64–75 | +2 | 0,6 | 0,6 | 1,1 | 89 % | 1,98 | 67 % |
| 4 | Gemini 2.5 Pro | 56 | 46–66 | -1 | 0,1 | 0,1 | 0,3 | 33 % | 1,14 | 71 % |
Law · 8 Items
Ranks 1–2 statistically tied (95 % CI overlaps rank 1)
In Law, Claude Opus 4.7 (top score) contradicts a stored key fact on average 0,1 times per answer and adds avg 0,7 claims the judge panel flags as false or unsupported; 75 % 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: 1,01 per 1000 output tokens, lowest GPT-5 at 0,80.
| # | Model | top100 | 95 % CI | Δ prev. | contradicts key fact | extra false/unsupp. | avg total | answers affected | per 1000 tok. | IRR |
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Claude Opus 4.7 | 78 | 68–86 | -7 | 0,1 | 0,7 | 0,8 | 75 % | 1,01 | 74 % |
| 2 | GPT-5 | 60 | 47–69 | -12 | 0,6 | 0,8 | 1,5 | 88 % | 0,80 | 68 % |
| 3 | Gemini 2.5 Pro | 55 | 45–65 | -5 | 0,5 | 0,1 | 0,4 | 50 % | 1,81 | 70 % |
| 4 | Mistral Large 2 | 48 | 44–52 | -8 | 0,5 | 0,8 | 1,3 | 88 % | 3,75 | 58 % |
Business law · 8 Items
Ranks 1–2 statistically tied (95 % CI overlaps rank 1)
In Business law, Claude Opus 4.7 (top score) contradicts a stored key fact on average 0,8 times per answer and adds avg 2,4 claims the judge panel flags as false or unsupported; 100 % of answers carry at least one flag. The lowest value is GPT-5 at avg 1,4 per answer. Relative to answer length: 1,71 per 1000 output tokens, lowest GPT-5 at 0,49.
| # | Model | top100 | 95 % CI | Δ prev. | contradicts key fact | extra false/unsupp. | avg total | answers affected | per 1000 tok. | IRR |
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Claude Opus 4.7 | 75 | 65–86 | -2 | 0,8 | 2,4 | 3,2 | 100 % | 1,71 | 79 % |
| 2 | GPT-5 | 75 | 64–85 | -9 | 0,5 | 0,9 | 1,4 | 100 % | 0,49 | 64 % |
| 3 | Mistral Large 2 | 52 | 39–63 | -1 | 1,1 | 2,0 | 3,2 | 100 % | 3,93 | 50 % |
| 4 | Gemini 2.5 Pro | 50 | 40–58 | +5 | 1,1 | 0,7 | 1,6 | 100 % | 4,29 | 57 % |
Raw data of this run
raw.jsonl and progress.jsonl are the public streams with response texts stripped (contamination protection, methodology §7).