Leaderboard

Models on the fully-open question set . Prompt and ground truth are public, so anyone can reproduce the runs.

BoardHeld-outPracticeReference

Edition v1 · cut 7/7/2026 · 17 models · 22 questions · 22 cells imputed (worst-of-others)

Per-model aggregate

Mean Cramér-log is the ranking metric (lower is better), scored against the verified ground-truth distribution. The Answered column shows real vs imputed; a malformed or unscorable answer counts as unanswered.

#ModelMean CramérAnsweredRan
1deepseek/deepseek-v4-pro0.166322/227/7/2026
2anthropic/claude-opus-4.80.255822/227/7/2026
3google/gemini-3.1-pro-preview0.285022/227/7/2026
4openai/gpt-5.50.306422/227/7/2026
5openai/gpt-5-mini0.370121/22 · 1 imp7/7/2026
6z-ai/glm-5.20.372521/22 · 1 imp7/7/2026
7deepseek/deepseek-v3.20.383721/22 · 1 imp7/7/2026
8openai/o30.438621/22 · 1 imp7/7/2026
9anthropic/claude-sonnet-4.60.459322/227/7/2026
10anthropic/claude-sonnet-50.604922/227/7/2026
11openai/o3-mini0.637422/227/7/2026
12x-ai/grok-4.30.651922/227/7/2026
13anthropic/claude-fable-50.684821/22 · 1 imp7/8/2026
14~anthropic/claude-fable-latest0.729520/22 · 2 imp7/8/2026
15qwen/qwen3-235b-a22b-thinking-25070.803821/22 · 1 imp7/7/2026
16openai/gpt-5.5-pro1.22319/22 · 13 imp7/9/2026
17meta-llama/llama-4-maverick1.780621/22 · 1 imp7/7/2026

Per-question Cramér-log heatmap

Each cell is one model's Cramér-log on that question. Imputed cells (the model didn't validly answer) show the worst-of-others penalty, marked .

Questiondeepseek-v4-proclaude-opus-4.8gemini-3.1-pro-previewgpt-5.5gpt-5-miniglm-5.2deepseek-v3.2o3claude-sonnet-4.6claude-sonnet-5o3-minigrok-4.3claude-fable-5claude-fable-latestqwen3-235b-a22b-thinking-2507gpt-5.5-prollama-4-maverick
annual-concrete-grand-canyon-fraction0.0900.0530.0240.0830.0780.0620.0430.2241.1700.0530.1070.0890.021🏆0.0320.0500.0550.742
bangladesh-tshirt-area-vs-factories0.4460.2230.0400.1020.0440.4320.2200.1300.020🏆0.0990.6471.0320.4000.2110.1181.0320.912
california-almond-water-per-dollar0.6100.051🏆0.2790.2660.5800.5021.4510.1571.1790.2571.4510.3950.2430.3290.5311.4511.031
chaco-soybean-crossover0.0320.0811.0110.3570.3060.0520.7061.1880.4750.5760.1490.0560.7791.1880.008🏆1.1880.881
china-concrete-three-gorges0.0220.0560.0130.1400.0670.2190.0480.7410.0220.0290.1630.004🏆0.0660.0960.1380.0870.741
electricity-vs-soy-per-worker0.1390.034🏆0.7650.5731.3370.1810.7750.1531.4230.9910.9690.2830.8570.8460.7350.3250.054
germany-coal-to-solar-land-fraction0.064🏆1.4460.5610.2630.7622.9170.0840.2601.3880.5910.3780.3430.3850.9352.9172.9172.741
honduras-costa-rica-homicide-gap0.539🏆0.8900.7270.5611.1631.0831.0290.9571.0930.5971.2730.8130.9380.9340.5720.6191.273
india-solid-fuel-to-lpg-import-years0.0400.2450.0340.1240.0370.2440.0200.0590.0860.4341.5030.0590.0390.001🏆0.3780.0687.599
nigeria-korea-cereal-yields0.3480.2980.025🏆0.1570.2120.0510.0840.0730.1320.2620.0530.0520.5760.3930.9620.9620.173
ocean-fish-biomass0.027🏆0.2050.9011.1060.0660.0280.2050.7530.5814.8200.1541.9438.2388.2388.2380.7988.238
paraguay-electricity-per-co20.0320.0240.0470.0260.3480.005🏆0.0620.0800.0530.1110.8990.0870.0220.0250.0591.5321.532
paraguay-itaipu-brazil-fraction0.0430.2020.0220.012🏆0.1190.0650.7560.5660.0920.6901.0200.8410.8710.7490.2451.0200.221
philippines-remittances-vs-timor-leste0.3730.0930.1830.0800.1590.1410.1190.0800.1000.0540.1610.1200.1190.1130.0760.048🏆1.948
poland-air-quality-coal-miner-wage-years0.0560.0500.0990.2640.1400.041🏆0.2291.4260.2370.9660.7240.0660.4120.4710.1051.4260.850
saudi-desal-per-oil-barrel0.1570.1710.0950.1170.5320.4470.1740.5320.5510.2870.7210.4240.0560.0450.025🏆0.7210.319
saudi-oil-solar-replacement-multiple0.1560.029🏆0.8131.4540.4900.0330.0580.3090.3910.2470.3527.1100.1230.2770.1971.3880.237
ssa-solar-ethiopia-gdp-years0.1321.0030.1480.015🏆0.0771.1641.1090.7180.0370.9261.8800.0330.0270.2661.2721.8801.109
tokyo-car-commute-lane-km0.003🏆0.2740.0100.7351.0630.2601.1400.8960.3961.2630.3150.4990.4530.4830.1081.2630.763
uruguay-share-coal-units-shutdown0.0630.0280.2980.0540.0350.1290.0350.005🏆0.1540.0130.3070.0180.0320.0400.5437.6707.670
vietnam-vs-myanmar-rice-per-worker0.0340.1370.0390.0750.1410.1000.0230.0320.3660.0270.7030.0680.1980.1530.021🏆0.0730.130
world-paved-roads0.2550.0350.1360.1740.3850.0410.0710.3090.1590.0140.0950.008🏆0.2090.2220.3850.3850.012
< 0.05 tight< 0.2 clean< 0.7 productive< 2 different interpretation≥ 2 suspect† imputed (worst-of-others; not a real answer)