Benchmarks
Intelligence test
Models are ranked by pairwise comparison, not by an absolute score. On every prompt each pair of models is compared head-to-head, and a Bradley-Terry model fits a rating from those comparisons. The win rate below turns that rating into the question it answers: how often this model beats an average model on this board. 50% is average.
8,603
scored runs
117,970
pairwise comparisons
24
prompts
4,699
runs without a score (provider error)
22 of 24 prompts separate models at all. The rest are answered equally well by nearly everyone and carry no ranking signal — that is what an absolute score near 100 was really measuring.
- Coding 4/4
- Creative 2/2
- Factual 3/4
- Multilingual 7/8
- Reasoning 4/4
- Healthcare 2/2
Rank 1OpenAI
73%±2
Win rate
Elo 1175
gpt-4.1-mini
Tier C
- Won / lost / tied
- 1296 / 172 / 1402
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 96.0
By category
Coding80%
Creative91%
Factual74%
Multilingual60%
Reasoning77%
Healthcare75%
Rank 1OpenAI
73%±2
Win rate
Elo 1172
gpt-5.1
Tier B
- Won / lost / tied
- 1125 / 101 / 1443
- prompts
- 20
- comparisons
- 2669
- Judge mean
- 95.0
By category
Coding76%
Creative79%
Factual71%
Multilingual70%
Reasoning78%
Rank 1OpenAI
73%±2
Win rate
Elo 1171
gpt-5.5-2026-04-23
Tier A
- Won / lost / tied
- 1149 / 132 / 1388
- prompts
- 20
- comparisons
- 2669
- Judge mean
- 95.9
By category
Coding78%
Creative89%
Factual76%
Multilingual67%
Reasoning66%
Rank 1OpenAI
70%±2
Win rate
Elo 1149
gpt-5.5
Tier C
- Won / lost / tied
- 1074 / 193 / 1402
- prompts
- 20
- comparisons
- 2669
- Judge mean
- 93.9
By category
Coding70%
Creative82%
Factual72%
Multilingual66%
Reasoning74%
Rank 1OpenAI
70%±2
Win rate
Elo 1149
gpt-4.1-2025-04-14
Tier C
- Won / lost / tied
- 1142 / 193 / 1535
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 96.0
By category
Coding72%
Creative92%
Factual72%
Multilingual65%
Reasoning66%
Healthcare84%
Rank 2OpenAI
70%±2
Win rate
Elo 1144
gpt-5.2-2025-12-11
Tier B
- Won / lost / tied
- 1062 / 216 / 1391
- prompts
- 20
- comparisons
- 2669
- Judge mean
- 93.9
By category
Coding75%
Creative78%
Factual57%
Multilingual69%
Reasoning78%
Rank 3OpenAI
69%±2
Win rate
Elo 1142
gpt-5.4
Tier A
- Won / lost / tied
- 1047 / 214 / 1408
- prompts
- 20
- comparisons
- 2669
- Judge mean
- 94.8
By category
Coding68%
Creative90%
Factual63%
Multilingual68%
Reasoning74%
Rank 4OpenAI
69%±2
Win rate
Elo 1141
gpt-5.2
Tier B
- Won / lost / tied
- 1062 / 235 / 1372
- prompts
- 20
- comparisons
- 2669
- Judge mean
- 94.2
By category
Coding66%
Creative92%
Factual67%
Multilingual66%
Reasoning74%
Rank 4OpenAI
69%±2
Win rate
Elo 1139
gpt-4.1
Tier B
- Won / lost / tied
- 1128 / 249 / 1493
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 95.7
By category
Coding74%
Creative83%
Factual69%
Multilingual64%
Reasoning66%
Healthcare79%
Rank 4OpenAI
69%±2
Win rate
Elo 1139
gpt-4.1-mini-2025-04-14
Tier C
- Won / lost / tied
- 1099 / 220 / 1551
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 95.1
By category
Coding69%
Creative82%
Factual67%
Multilingual68%
Reasoning73%
Healthcare70%
Rank 4Anthropic
69%±2
Win rate
Elo 1135
Claude Opus 4.6
Tier B
- Won / lost / tied
- 1162 / 309 / 1399
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 97.0
By category
Coding63%
Creative93%
Factual73%
Multilingual64%
Reasoning67%
Healthcare90%
Rank 4OpenAI
68%±2
Win rate
Elo 1133
gpt-5.4-mini
Tier A
- Won / lost / tied
- 1067 / 292 / 1310
- prompts
- 20
- comparisons
- 2669
- Judge mean
- 94.1
By category
Coding71%
Creative92%
Factual61%
Multilingual63%
Reasoning71%
Rank 4Z.ai (GLM / Zhipu)
68%±2
Win rate
Elo 1132
GLM-4.7
Tier A
- Won / lost / tied
- 838 / 259 / 882
- prompts
- 14
- comparisons
- 1979
- Judge mean
- 91.1
By category
Coding61%
Creative85%
Factual70%
Reasoning74%
Rank 4OpenAI
68%±2
Win rate
Elo 1129
gpt-5.4-2026-03-05
Tier B
- Won / lost / tied
- 1035 / 289 / 1345
- prompts
- 20
- comparisons
- 2669
- Judge mean
- 94.6
By category
Coding74%
Creative89%
Factual55%
Multilingual64%
Reasoning74%
Rank 4OpenAI
67%±2
Win rate
Elo 1127
gpt-5.1-2025-11-13
Tier B
- Won / lost / tied
- 987 / 252 / 1430
- prompts
- 20
- comparisons
- 2669
- Judge mean
- 94.0
By category
Coding75%
Creative73%
Factual59%
Multilingual64%
Reasoning74%
Rank 6OpenAI
66%±2
Win rate
Elo 1117
gpt-5.4-mini-2026-03-17
Tier A
- Won / lost / tied
- 981 / 315 / 1373
- prompts
- 20
- comparisons
- 2669
- Judge mean
- 93.0
By category
Coding65%
Creative89%
Factual59%
Multilingual63%
Reasoning74%
Rank 6Z.ai (GLM / Zhipu)
66%±2
Win rate
Elo 1115
GLM-4.5
Tier A
- Won / lost / tied
- 843 / 337 / 862
- prompts
- 14
- comparisons
- 2042
- Judge mean
- 90.0
By category
Coding71%
Creative85%
Factual66%
Reasoning55%
Rank 12OpenAI
65%±2
Win rate
Elo 1106
gpt-4o
Tier C
- Won / lost / tied
- 1034 / 390 / 1446
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 94.8
By category
Coding58%
Creative73%
Factual68%
Multilingual64%
Reasoning73%
Healthcare67%
Rank 14OpenAI
64%±2
Win rate
Elo 1101
gpt-5-search-api
Tier C
- Won / lost / tied
- 1026 / 421 / 1423
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 94.5
By category
Coding66%
Creative86%
Factual61%
Multilingual53%
Reasoning70%
Healthcare87%
Rank 14OpenAI
64%±2
Win rate
Elo 1101
- Won / lost / tied
- 1060 / 455 / 1355
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 94.0
By category
Coding62%
Creative82%
Factual62%
Multilingual52%
Reasoning77%
Healthcare89%
Rank 14OpenAI
64%±2
Win rate
Elo 1101
gpt-4o-2024-05-13
Tier C
- Won / lost / tied
- 1004 / 401 / 1465
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 95.2
By category
Coding62%
Creative82%
Factual62%
Multilingual64%
Reasoning66%
Healthcare75%
Rank 15Anthropic
64%±2
Win rate
Elo 1100
Claude Opus 4.7
Tier B
- Won / lost / tied
- 1038 / 437 / 1395
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 95.1
By category
Coding51%
Creative60%
Factual79%
Multilingual60%
Reasoning73%
Healthcare81%
Rank 16Google Gemini
64%±2
Win rate
Elo 1098
Gemini Flash-Lite Latest
Tier C
- Won / lost / tied
- 1069 / 483 / 1318
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 94.8
By category
Coding67%
Creative45%
Factual71%
Multilingual61%
Reasoning64%
Healthcare85%
Rank 16Anthropic
63%±2
Win rate
Elo 1094
Claude Fable 5
Tier A
- Won / lost / tied
- 919 / 401 / 1390
- prompts
- 20
- comparisons
- 2710
- Judge mean
- 88.9
By category
Coding64%
Creative67%
Factual59%
Multilingual65%
Reasoning69%
Rank 16OpenAI
63%±2
Win rate
Elo 1092
o1
Tier C
- Won / lost / tied
- 879 / 398 / 1262
- prompts
- 20
- comparisons
- 2539
- Judge mean
- 90.4
By category
Coding77%
Creative55%
Factual48%
Multilingual65%
Reasoning63%
Rank 16Anthropic
63%±2
Win rate
Elo 1090
Claude Opus 4.5
Tier B
- Won / lost / tied
- 1008 / 484 / 1378
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 95.8
By category
Coding54%
Creative58%
Factual71%
Multilingual63%
Reasoning67%
Healthcare88%
Rank 16Z.ai (GLM / Zhipu)
63%±2
Win rate
Elo 1089
GLM-5
Tier A
- Won / lost / tied
- 553 / 252 / 733
- prompts
- 14
- comparisons
- 1538
- Judge mean
- 94.6
By category
Coding57%
Creative89%
Factual64%
Reasoning61%
Rank 16Google Gemini
62%±2
Win rate
Elo 1087
Nano Banana
Tier B
- Won / lost / tied
- 615 / 274 / 910
- prompts
- 23
- comparisons
- 1799
- Judge mean
- 93.5
By category
Coding51%
Creative70%
Factual61%
Multilingual65%
Reasoning73%
Healthcare58%
Rank 17OpenAI
62%±2
Win rate
Elo 1086
o1-2024-12-17
Tier C
- Won / lost / tied
- 826 / 396 / 1246
- prompts
- 20
- comparisons
- 2468
- Judge mean
- 86.8
By category
Coding77%
Creative6%
Factual61%
Multilingual68%
Reasoning63%
Rank 17OVH AI Endpoints (GRA)
62%±2
Win rate
Elo 1085
gpt-oss-120b
Tier C
- Won / lost / tied
- 901 / 450 / 1318
- prompts
- 20
- comparisons
- 2669
- Judge mean
- 92.0
By category
Coding60%
Creative84%
Factual61%
Multilingual54%
Reasoning70%
Rank 18OpenAI
62%±2
Win rate
Elo 1083
gpt-4o-2024-08-06
Tier C
- Won / lost / tied
- 987 / 519 / 1364
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 92.8
By category
Coding51%
Creative73%
Factual72%
Multilingual62%
Reasoning62%
Healthcare72%
Rank 18OpenAI
62%±2
Win rate
Elo 1082
gpt-4o-2024-11-20
Tier C
- Won / lost / tied
- 940 / 479 / 1451
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 94.3
By category
Coding53%
Creative50%
Factual71%
Multilingual58%
Reasoning71%
Healthcare86%
Rank 18Google Gemini
61%±2
Win rate
Elo 1079
Gemini 3.1 Flash Lite
Tier B
- Won / lost / tied
- 853 / 444 / 1372
- prompts
- 20
- comparisons
- 2669
- Judge mean
- 92.8
By category
Coding55%
Creative42%
Factual70%
Multilingual59%
Reasoning74%
Rank 19OpenAI
61%±2
Win rate
Elo 1076
o3
Tier C
- Won / lost / tied
- 792 / 432 / 1169
- prompts
- 20
- comparisons
- 2393
- Judge mean
- 83.3
By category
Coding65%
Creative88%
Factual57%
Multilingual53%
Reasoning58%
Rank 19Anthropic
61%±2
Win rate
Elo 1075
Claude Sonnet 4.6
Tier A
- Won / lost / tied
- 935 / 521 / 1414
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 94.5
By category
Coding56%
Creative54%
Factual64%
Multilingual60%
Reasoning67%
Healthcare82%
Rank 19OpenAI
61%±2
Win rate
Elo 1075
gpt-4o-mini
Tier C
- Won / lost / tied
- 927 / 516 / 1427
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 89.1
By category
Coding60%
Creative72%
Factual46%
Multilingual67%
Reasoning69%
Healthcare54%
Rank 19OpenAI
61%±2
Win rate
Elo 1074
gpt-4.1-nano-2025-04-14
Tier C
- Won / lost / tied
- 926 / 520 / 1424
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 92.3
By category
Coding66%
Creative66%
Factual59%
Multilingual54%
Reasoning69%
Healthcare35%
Rank 19OpenAI
60%±2
Win rate
Elo 1074
o4-mini-2025-04-16
Tier B
- Won / lost / tied
- 788 / 437 / 1239
- prompts
- 20
- comparisons
- 2464
- Judge mean
- 81.6
By category
Coding58%
Creative69%
Factual50%
Multilingual70%
Reasoning63%
Rank 23Anthropic
60%±2
Win rate
Elo 1073
Claude Sonnet 4.5
Tier B
- Won / lost / tied
- 961 / 564 / 1345
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 94.0
By category
Coding40%
Creative57%
Factual70%
Multilingual61%
Reasoning77%
Healthcare78%
Rank 22OpenAI
60%±2
Win rate
Elo 1073
o4-mini
Tier C
- Won / lost / tied
- 850 / 497 / 1192
- prompts
- 20
- comparisons
- 2539
- Judge mean
- 81.3
By category
Coding71%
Creative79%
Factual54%
Multilingual57%
Reasoning49%
Rank 23OpenAI
60%±2
Win rate
Elo 1072
gpt-4-turbo-2024-04-09
Tier C
- Won / lost / tied
- 965 / 573 / 1332
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 93.6
By category
Coding56%
Creative77%
Factual57%
Multilingual55%
Reasoning73%
Healthcare72%
Rank 19Google Gemini
60%±3
Win rate
Elo 1070
Nano Banana 2
Tier B
- Won / lost / tied
- 501 / 290 / 608
- prompts
- 18
- comparisons
- 1399
- Judge mean
- 92.6
By category
Coding70%
Creative61%
Factual53%
Multilingual67%
Reasoning64%
Healthcare36%
Rank 23Anthropic
60%±2
Win rate
Elo 1069
Claude Sonnet 5
Tier A
- Won / lost / tied
- 598 / 400 / 700
- prompts
- 17
- comparisons
- 1698
- Judge mean
- 86.9
By category
Coding54%
Creative89%
Factual53%
Multilingual65%
Reasoning59%
Rank 24OpenAI
60%±2
Win rate
Elo 1068
o3-2025-04-16
Tier B
- Won / lost / tied
- 714 / 414 / 1246
- prompts
- 19
- comparisons
- 2374
- Judge mean
- 82.1
By category
Coding65%
Creative84%
Factual49%
Multilingual56%
Reasoning58%
Rank 25OpenAI
59%±2
Win rate
Elo 1064
o3-mini-2025-01-31
Tier C
- Won / lost / tied
- 817 / 516 / 1269
- prompts
- 20
- comparisons
- 2602
- Judge mean
- 90.5
By category
Coding67%
Creative72%
Factual41%
Multilingual64%
Reasoning60%
Rank 25OpenAI
59%±2
Win rate
Elo 1064
gpt-5.4-nano-2026-03-17
Tier A
- Won / lost / tied
- 848 / 544 / 1277
- prompts
- 20
- comparisons
- 2669
- Judge mean
- 85.2
By category
Coding63%
Creative30%
Factual44%
Multilingual69%
Reasoning74%
Rank 27OpenAI
59%±2
Win rate
Elo 1060
gpt-4-turbo
Tier C
- Won / lost / tied
- 899 / 601 / 1370
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 93.4
By category
Coding49%
Creative71%
Factual69%
Multilingual56%
Reasoning59%
Healthcare78%
Rank 27Anthropic
58%±2
Win rate
Elo 1058
Claude Opus 4.8
Tier A
- Won / lost / tied
- 840 / 578 / 1251
- prompts
- 20
- comparisons
- 2669
- Judge mean
- 93.8
By category
Coding48%
Creative44%
Factual72%
Multilingual58%
Reasoning67%
Rank 34OpenAI
57%±2
Win rate
Elo 1052
gpt-4.1-nano
Tier C
- Won / lost / tied
- 876 / 641 / 1353
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 91.1
By category
Coding56%
Creative60%
Factual64%
Multilingual53%
Reasoning65%
Healthcare21%
Rank 34OpenAI
57%±2
Win rate
Elo 1050
gpt-4o-mini-2024-07-18
Tier C
- Won / lost / tied
- 904 / 681 / 1285
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 88.5
By category
Coding61%
Creative70%
Factual44%
Multilingual58%
Reasoning65%
Healthcare49%
Rank 34OVH AI Endpoints (GRA)
57%±2
Win rate
Elo 1048
- Won / lost / tied
- 778 / 584 / 1307
- prompts
- 20
- comparisons
- 2669
- Judge mean
- 91.9
By category
Coding46%
Creative69%
Factual60%
Multilingual51%
Reasoning74%
Rank 47Anthropic
55%±2
Win rate
Elo 1035
Claude Haiku 4.5
Tier A
- Won / lost / tied
- 862 / 754 / 1254
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 91.7
By category
Coding58%
Creative55%
Factual53%
Multilingual55%
Reasoning61%
Healthcare33%
Rank 47OVH AI Endpoints (GRA)
55%±2
Win rate
Elo 1033
- Won / lost / tied
- 828 / 744 / 1097
- prompts
- 20
- comparisons
- 2669
- Judge mean
- 89.2
By category
Coding55%
Creative36%
Factual57%
Multilingual53%
Reasoning64%
Rank 49OVH AI Endpoints (GRA)
54%±2
Win rate
Elo 1027
- Won / lost / tied
- 792 / 746 / 1131
- prompts
- 20
- comparisons
- 2669
- Judge mean
- 90.9
By category
Coding66%
Creative41%
Factual46%
Multilingual50%
Reasoning61%
Rank 50OVH AI Endpoints (GRA)
53%±2
Win rate
Elo 1024
Qwen2.5-VL-72B-Instruct
Tier B
- Won / lost / tied
- 791 / 767 / 1111
- prompts
- 20
- comparisons
- 2669
- Judge mean
- 91.5
By category
Coding56%
Creative65%
Factual33%
Multilingual52%
Reasoning74%
Rank 52OpenAI
53%±2
Win rate
Elo 1019
gpt-5.4-nano
Tier C
- Won / lost / tied
- 835 / 845 / 989
- prompts
- 20
- comparisons
- 2669
- Judge mean
- 84.3
By category
Coding55%
Creative32%
Factual32%
Multilingual68%
Reasoning66%
Rank 52OpenAI
51%±2
Win rate
Elo 1009
gpt-4
Tier C
- Won / lost / tied
- 867 / 957 / 1046
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 88.6
By category
Coding58%
Creative33%
Factual41%
Multilingual54%
Reasoning60%
Healthcare43%
Rank 53OpenAI
51%±2
Win rate
Elo 1005
o3-mini
Tier C
- Won / lost / tied
- 627 / 722 / 1119
- prompts
- 20
- comparisons
- 2468
- Judge mean
- 84.8
By category
Coding58%
Creative30%
Factual31%
Multilingual57%
Reasoning64%
Rank 53Google Gemini
50%±2
Win rate
Elo 1003
Gemini Robotics-ER 1.6 Preview
Tier Barchived
- Won / lost / tied
- 485 / 536 / 778
- prompts
- 23
- comparisons
- 1799
- Judge mean
- 79.3
By category
Coding44%
Creative82%
Factual43%
Multilingual55%
Reasoning43%
Healthcare43%
Rank 54OpenAI
50%±2
Win rate
Elo 1003
gpt-5-mini
Tier C
- Won / lost / tied
- 549 / 641 / 1130
- prompts
- 20
- comparisons
- 2320
- Judge mean
- 75.1
By category
Coding60%
Creative50%
Factual40%
Multilingual45%
Reasoning58%
Rank 54Anthropic
50%±2
Win rate
Elo 999
Claude Opus 5
Tier A
- Won / lost / tied
- 586 / 716 / 868
- prompts
- 18
- comparisons
- 2170
- Judge mean
- 82.7
By category
Coding31%
Creative86%
Factual53%
Multilingual41%
Reasoning55%
Rank 56Google Gemini
49%±2
Win rate
Elo 996
Gemini 2.5 Flash-Lite
Tier B
- Won / lost / tied
- 801 / 988 / 1081
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 90.7
By category
Coding38%
Creative51%
Factual45%
Multilingual57%
Reasoning60%
Healthcare53%
Rank 59OpenAI
47%±2
Win rate
Elo 977
gpt-3.5-turbo-1106
Tier C
- Won / lost / tied
- 730 / 1061 / 1079
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 82.2
By category
Coding55%
Creative54%
Factual23%
Multilingual56%
Reasoning52%
Healthcare18%
Rank 59OpenAI
47%±2
Win rate
Elo 977
gpt-4-0613
Tier Carchived
- Won / lost / tied
- 754 / 1085 / 1031
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 90.0
By category
Coding45%
Creative53%
Factual37%
Multilingual48%
Reasoning56%
Healthcare51%
Rank 58OpenAI
47%±2
Win rate
Elo 976
gpt-5-mini-2025-08-07
Tier B
- Won / lost / tied
- 549 / 818 / 1026
- prompts
- 20
- comparisons
- 2393
- Judge mean
- 71.9
By category
Coding64%
Creative35%
Factual40%
Multilingual32%
Reasoning58%
Rank 63OpenAI
43%±2
Win rate
Elo 952
gpt-3.5-turbo-0125
Tier C
- Won / lost / tied
- 691 / 1203 / 976
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 82.4
By category
Coding58%
Creative46%
Factual18%
Multilingual51%
Reasoning43%
Healthcare25%
Rank 66Google Gemini
42%±2
Win rate
Elo 944
Gemini 3 Flash Preview
Tier C
- Won / lost / tied
- 686 / 1258 / 926
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 77.5
By category
Coding35%
Creative21%
Factual52%
Multilingual42%
Reasoning50%
Healthcare23%
Rank 66OpenAI
42%±2
Win rate
Elo 942
gpt-5
Tier C
- Won / lost / tied
- 407 / 813 / 861
- prompts
- 19
- comparisons
- 2081
- Judge mean
- 62.0
By category
Coding56%
Creative7%
Factual39%
Multilingual38%
Reasoning45%
Rank 66OpenAI
41%±2
Win rate
Elo 939
gpt-3.5-turbo
Tier C
- Won / lost / tied
- 688 / 1291 / 891
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 82.3
By category
Coding49%
Creative54%
Factual18%
Multilingual43%
Reasoning53%
Healthcare22%
Rank 66OpenAI
41%±2
Win rate
Elo 936
gpt-3.5-turbo-16k
Tier C
- Won / lost / tied
- 681 / 1307 / 882
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 81.1
By category
Coding63%
Creative30%
Factual21%
Multilingual47%
Reasoning33%
Healthcare20%
Rank 66OpenAI
40%±2
Win rate
Elo 933
gpt-5-2025-08-07
Tier B
- Won / lost / tied
- 466 / 977 / 877
- prompts
- 20
- comparisons
- 2320
- Judge mean
- 59.1
By category
Coding56%
Creative3%
Factual24%
Multilingual37%
Reasoning58%
Rank 67OVH AI Endpoints (GRA)
39%±2
Win rate
Elo 922
gpt-oss-20b
Tier C
- Won / lost / tied
- 559 / 1195 / 781
- prompts
- 20
- comparisons
- 2535
- Judge mean
- 79.8
By category
Coding50%
Creative45%
Factual31%
Multilingual25%
Reasoning49%
Rank 68Z.ai (GLM / Zhipu)
38%±2
Win rate
Elo 912
GLM-5 Turbo
Tier B
- Won / lost / tied
- 358 / 847 / 611
- prompts
- 14
- comparisons
- 1816
- Judge mean
- 54.5
By category
Coding43%
Creative31%
Factual44%
Reasoning24%
Rank 68Z.ai (GLM / Zhipu)
37%±2
Win rate
Elo 909
GLM-5.2
Tier A
- Won / lost / tied
- 367 / 846 / 538
- prompts
- 14
- comparisons
- 1751
- Judge mean
- 54.0
By category
Coding50%
Creative4%
Factual42%
Reasoning20%
Rank 70Z.ai (GLM / Zhipu)
37%±2
Win rate
Elo 907
GLM-4.5 Air
Tier B
- Won / lost / tied
- 367 / 881 / 578
- prompts
- 14
- comparisons
- 1826
- Judge mean
- 64.5
By category
Coding50%
Creative46%
Factual31%
Reasoning15%
Rank 71Z.ai (GLM / Zhipu)
37%±2
Win rate
Elo 904
GLM-5.1
Tier A
- Won / lost / tied
- 375 / 831 / 412
- prompts
- 14
- comparisons
- 1618
- Judge mean
- 69.1
By category
Coding32%
Creative13%
Factual50%
Reasoning34%
Rank 73OVH AI Endpoints (GRA)
35%±2
Win rate
Elo 890
- Won / lost / tied
- 555 / 1438 / 676
- prompts
- 20
- comparisons
- 2669
- Judge mean
- 75.9
By category
Coding42%
Creative10%
Factual29%
Multilingual39%
Reasoning36%
Rank 73Google Gemini
34%±2
Win rate
Elo 887
Gemini Flash Latest
Tier B
- Won / lost / tied
- 605 / 1574 / 691
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 63.9
By category
Coding32%
Creative31%
Factual49%
Multilingual29%
Reasoning28%
Healthcare9%
Rank 77Z.ai (GLM / Zhipu)
31%±2
Win rate
Elo 864
GLM-4.5V (vision)
Tier A
- Won / lost / tied
- 303 / 941 / 449
- prompts
- 14
- comparisons
- 1693
- Judge mean
- 54.4
By category
Coding50%
Creative8%
Factual39%
Reasoning6%
Rank 77Z.ai (GLM / Zhipu)
31%±2
Win rate
Elo 862
GLM-4.6
Tier A
- Won / lost / tied
- 303 / 974 / 463
- prompts
- 13
- comparisons
- 1740
- Judge mean
- 55.8
By category
Coding46%
Creative1%
Factual37%
Reasoning16%
Rank 78OpenAI
30%±2
Win rate
Elo 856
gpt-5-nano-2025-08-07
Tier B
- Won / lost / tied
- 268 / 990 / 568
- prompts
- 18
- comparisons
- 1826
- Judge mean
- 39.6
By category
Coding27%
Creative7%
Factual48%
Multilingual27%
Reasoning21%
Rank 79OVH AI Endpoints (GRA)
28%±2
Win rate
Elo 835
Mistral-7B-Instruct-v0.3
Tier C
- Won / lost / tied
- 476 / 1684 / 509
- prompts
- 20
- comparisons
- 2669
- Judge mean
- 71.7
By category
Coding35%
Creative11%
Factual27%
Multilingual27%
Reasoning24%
Rank 81Z.ai (GLM / Zhipu)
26%±2
Win rate
Elo 818
GLM-4.6V (vision)
Tier A
- Won / lost / tied
- 230 / 1025 / 421
- prompts
- 14
- comparisons
- 1676
- Judge mean
- 46.4
By category
Coding43%
Creative4%
Factual29%
Reasoning5%
Rank 82Google Gemini
25%±2
Win rate
Elo 812
Gemini 3.5 Flash
Tier A
- Won / lost / tied
- 418 / 1752 / 499
- prompts
- 20
- comparisons
- 2669
- Judge mean
- 50.5
By category
Coding9%
Creative48%
Factual43%
Multilingual25%
Reasoning10%
Rank 83OpenAI
24%±2
Win rate
Elo 799
gpt-5-nano
Tier C
- Won / lost / tied
- 219 / 1207 / 501
- prompts
- 20
- comparisons
- 1927
- Judge mean
- 38.0
By category
Coding18%
Creative4%
Factual50%
Multilingual18%
Reasoning10%
Rank 83Google Gemini
23%±2
Win rate
Elo 792
Nano Banana Pro
Tier A
- Won / lost / tied
- 192 / 915 / 292
- prompts
- 18
- comparisons
- 1399
- Judge mean
- 41.4
By category
Coding19%
Creative14%
Factual44%
Multilingual4%
Reasoning3%
Healthcare54%
Rank 83OVH AI Endpoints (GRA)
23%±2
Win rate
Elo 790
Qwen3.5-397B-A17B
Tier A
- Won / lost / tied
- 255 / 1313 / 428
- prompts
- 19
- comparisons
- 1996
- Judge mean
- 43.4
By category
Coding42%
Creative6%
Factual6%
Multilingual28%
Reasoning5%
Rank 83OVH AI Endpoints (GRA)
23%±2
Win rate
Elo 789
Qwen3-32B
Tier B
- Won / lost / tied
- 440 / 1893 / 336
- prompts
- 20
- comparisons
- 2669
- Judge mean
- 75.8
By category
Coding18%
Creative14%
Factual21%
Multilingual27%
Reasoning24%
Rank 83Google Gemini
22%±2
Win rate
Elo 780
- Won / lost / tied
- 250 / 1275 / 353
- prompts
- 24
- comparisons
- 1878
- Judge mean
- 44.1
By category
Coding4%
Creative38%
Factual42%
Multilingual19%
Reasoning13%
Healthcare3%
Rank 85OVH AI Endpoints (GRA)
21%±2
Win rate
Elo 770
Qwen3.5-9B
Tier B
- Won / lost / tied
- 211 / 1343 / 442
- prompts
- 19
- comparisons
- 1996
- Judge mean
- 44.1
By category
Coding36%
Creative5%
Factual11%
Multilingual24%
Reasoning3%
Rank 85Google Gemini
21%±2
Win rate
Elo 767
Gemini 3.1 Pro Preview
Tier C
- Won / lost / tied
- 404 / 2080 / 386
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 47.4
By category
Coding5%
Creative51%
Factual43%
Multilingual16%
Reasoning7%
Healthcare5%
Rank 86Google Gemini
20%±2
Win rate
Elo 761
Gemini Pro Latest
Tier C
- Won / lost / tied
- 371 / 2077 / 422
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 43.9
By category
Coding4%
Creative23%
Factual49%
Multilingual17%
Reasoning7%
Healthcare2%
Rank 92Google Gemini
17%±2
Win rate
Elo 726
Gemini 2.5 Pro
Tier A
- Won / lost / tied
- 182 / 1437 / 345
- prompts
- 24
- comparisons
- 1964
- Judge mean
- 30.7
By category
Coding17%
Creative3%
Factual34%
Multilingual8%
Reasoning3%
Healthcare54%
Rank 93Google Gemini
14%±1
Win rate
Elo 687
Gemini 2.5 Flash
Tier A
- Won / lost / tied
- 295 / 2327 / 248
- prompts
- 24
- comparisons
- 2870
- Judge mean
- 36.6
By category
Coding3%
Creative24%
Factual25%
Multilingual11%
Reasoning10%
Healthcare2%
94 models ranked from 117,970 pairwise comparisons across 24 prompts.
Provisional
Too little evidence to rank: needs at least 5 prompts and 30 comparisons.
ProvisionalZ.ai (GLM / Zhipu)
1%±1
Win rate
Elo 155
CogView-4
Tier B
- Won / lost / tied
- 0 / 361 / 6
- prompts
- 4
- comparisons
- 367
- Judge mean
- 0.0
By category
Coding1%
Creative<1% · thin