CulturalRiddles - Exact Match: leaderboard

Deterministic rule-based exact-match accuracy (%) on Cohere Labs Community's CulturalRiddles benchmark across 57 human-evaluated communities prior to human and LLM-jury review.

Metric: Rule-based exact-match accuracy (%, 57 communities). Source: huggingface.co. Saturation forecast: Around February 2027. 25 models tracked. Updated 20 Nov 2026.

Top models

#ModelSettingScoreOverall rank
1Gemini 3.1 Pro (Preview)48.5#101
2GPT-5.5Non-reasoning42.8#54
3DeepSeek V4 ProHigh36.6#198
4Claude Opus 4.7High34.2#85
5Gemini 2.5 Flash25.3#469
6DeepSeek V4 Flash21.5#267
7Mistral Large 320.6#835
8Gemma 4 31B (IT)20.4#354
9Qwen-SEA-LION-v4.5-27B-IT17.7#603
10Gemini 2.5 Flash Lite17.4#974
11Qwen 3.5 27B13.8#378
12Apertus-v1.5-70B9.5#1783
13EuroLLM-22B-Instruct-25127.6#2012
14Apertus-70B-Instruct-25097.1#2627
15Nova Lite6.4#2322

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=culturalriddles-exact-match · How It Works · Data refreshed daily, snapshot 2026-11-20.