Ranking run 2026-05-12
Claude Opus 4.7 leads with 78/100 across all domains. Tax law: Claude Opus 4.7 (67) · Medicine: Claude Opus 4.7 (82) · Law: Claude Opus 4.7 (85) · Business law: GPT-5 (84).
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 / 25 | 78 | 73–82 | +1 | 49 | 1,59 | — |
| 2 | GPT-5 | 0 | 31 / 50 | 72 | 65–79 | -1 | 64 | 0,66 | — |
| 3 | Gemini 2.5 Pro | 0 | 23 / 27 | 49 | 42–55 | +1 | 73 | 4,18 | — |
| 3 | Mistral Large 2 | 0 | 17 / 366 | 49 | 41–57 | +1 | 42 | 3,91 | — |
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 0,8 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 Gemini 2.5 Pro at avg 2,0 per answer. Relative to answer length: 2,01 per 1000 output tokens, lowest GPT-5 at 0,91.
| # | 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 | 67 | 57–75 | -4 | 0,8 | 2,0 | 2,8 | 100 % | 2,01 | 83 % |
| 2 | GPT-5 | 50 | 35–63 | -13 | 1,3 | 1,3 | 2,5 | 100 % | 0,91 | 65 % |
| 3 | Gemini 2.5 Pro | 31 | 18–46 | -1 | 1,3 | 0,7 | 2,0 | 88 % | 8,81 | 71 % |
| 4 | Mistral Large 2 | 17 | 9–25 | +5 | 2,4 | 1,5 | 3,9 | 100 % | 8,75 | 54 % |
Medicine · 9 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 0,7 times per answer and adds avg 0,9 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 0,5 per answer. Relative to answer length: 1,55 per 1000 output tokens, lowest GPT-5 at 0,76.
| # | 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 | 82 | 73–88 | -3 | 0,7 | 0,9 | 1,6 | 100 % | 1,55 | 82 % |
| 2 | GPT-5 | 82 | 75–89 | -3 | 0,3 | 0,7 | 1,0 | 100 % | 0,76 | 76 % |
| 3 | Mistral Large 2 | 67 | 59–74 | -3 | 0,4 | 0,9 | 1,3 | 100 % | 2,00 | 75 % |
| 4 | Gemini 2.5 Pro | 57 | 45–68 | +3 | 0,2 | 0,3 | 0,5 | 67 % | 2,04 | 63 % |
Law · 8 Items
In Law, Claude Opus 4.7 (top score) contradicts a stored key fact on average 0,0 times per answer and adds avg 0,8 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 0,1 per answer. Relative to answer length: 0,98 per 1000 output tokens, lowest GPT-5 at 0,35.
| # | 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 | 85 | 81–90 | +5 | 0,0 | 0,8 | 0,8 | 75 % | 0,98 | 89 % |
| 2 | GPT-5 | 72 | 63–81 | +8 | 0,1 | 0,5 | 0,6 | 75 % | 0,35 | 72 % |
| 3 | Gemini 2.5 Pro | 60 | 53–68 | +9 | 0,1 | 0,0 | 0,1 | 25 % | 0,38 | 72 % |
| 4 | Mistral Large 2 | 56 | 50–62 | 0 | 0,3 | 0,5 | 0,7 | 75 % | 2,09 | 59 % |
Business law · 8 Items
Ranks 1–2 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 1,3 claims the judge panel flags as false or unsupported; 100 % of answers carry at least one flag. Relative to answer length: 0,57 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 | 84 | 79–90 | +7 | 0,4 | 1,3 | 1,6 | 100 % | 0,57 | 81 % |
| 2 | Claude Opus 4.7 | 77 | 71–84 | +6 | 0,6 | 2,3 | 2,9 | 100 % | 1,59 | 76 % |
| 3 | Mistral Large 2 | 53 | 38–64 | +4 | 1,4 | 2,1 | 3,5 | 100 % | 3,80 | 56 % |
| 4 | Gemini 2.5 Pro | 45 | 32–58 | -8 | 1,4 | 0,5 | 1,8 | 100 % | 5,23 | 60 % |
Raw data of this run
raw.jsonl and progress.jsonl are the public streams with response texts stripped (contamination protection, methodology §7).