Benchmarks · knowledge
PrinzBench
A lawyer's private benchmark: 25 legal research questions in a niche area of U.S. law (find the statutes, regulations and guidance online, then reach the right conclusion) and 8 needle-in-the-haystack web search questions. Each is asked three times and hand-graded by the author at pass@1, not blind; score out of 99. The author considers it saturated since June 2026 and no longer tests OpenAI's Pro models.
27 models · 27 results · updated 2026-07-18 · github.com ↗
Leaderboard
GPT-5.6 Sol Pro max
More from this benchmark
2 published measuresScore vs any published measure2 measures: sub-scores, splits, counts
Breakdown
darker is better within each column · click a column to sort| Model | ||
|---|---|---|
| GPT-5.6 Sol Pro max | 67.0 | 24.0 |
| GPT-5.5 Pro xhigh | 63.0 | 19.0 |
| GPT-5.4 Pro xhigh | 59.0 | 20.0 |
| GPT-5.5 xhigh | 56.0 | 18.0 |
| GPT-5.4 xhigh | 50.0 | 19.0 |
| Fable 5 max | 49.0 | 7 |
| GPT-5.2 xhigh | 41.0 | 11.0 |
| GPT-5.3 Codex high | 41.0 | 11.0 |
| Gemini 3.1 Pro high | 41.0 | 9 |
| Kimi K3 max | 36.0 | 11.0 |
| Grok 4.20 xhigh | 34.0 | 9 |
| Opus 4.8 max | 35.0 | 7 |
| Gemini 3 Flash high | 29.0 | 7 |
| Gemini 3 Pro high | 29.0 | 6 |
| Kimi K2.5 max | 28.0 | 7 |
| Muse Spark | 24.0 | 7 |
| GLM 5.2 max | 25.0 | 5 |
| Opus 4.7 max | 21.0 | 4 |
| Grok 4.1 | 19.0 | 6 |
| Qwen 3 Max | 18.0 | 7 |
| DeepSeek V4 Pro max | 17.0 | 6 |
| Grok 4 high | 16.0 | 7 |
| Kimi K2 medium | 18.0 | 4 |
| Sonnet 4.5 high | 20.0 | 0 |
| Opus 4.6 max | 18.0 | 1 |
All results
27 rows| # | Model | Effort | PrinzBench |
|---|---|---|---|
| 1 | GPT-5.6 Sol Pro | max | 91.0 |
| 2 | GPT-5.5 Pro | xhigh | 82.0 |
| 3 | GPT-5.4 Pro | xhigh | 79.0 |
| 4 | GPT-5.5 | xhigh | 74.0 |
| 5 | GPT-5.4 | xhigh | 69.0 |
| 6 | Fable 5 | max | 56.0 |
| 7 | GPT-5.2 | xhigh | 52.0 |
| 8 | GPT-5.3 Codex | high | 52.0 |
| 9 | Gemini 3.1 Pro | high | 50.0 |
| 10 | Kimi K3 | max | 47.0 |
| 11 | Grok 4.20 | xhigh | 43.0 |
| 12 | Opus 4.8 | max | 42.0 |
| 13 | Gemini 3 Flash | high | 36.0 |
| 14 | Gemini 3 Pro | high | 35.0 |
| 15 | Kimi K2.5 | max | 35.0 |
| 16 | Muse Spark | 31.0 | |
| 17 | GLM 5.2 | max | 30.0 |
| 18 | Opus 4.7 | max | 25.0 |
| 19 | Grok 4.1 | 25.0 | |
| 20 | Qwen 3 Max | max | 25.0 |
| 21 | DeepSeek V4 Pro | max | 23.0 |
| 22 | Grok 4 | high | 23.0 |
| 23 | Kimi K2 | medium | 22.0 |
| 24 | Sonnet 4.5 | high | 20.0 |
| 25 | Opus 4.6 | max | 19.0 |
| 26 | Opus 4.5 | high | 14.0 |
| 27 | Sonnet 4.6 | max | 8.0 |