MSQA - Beliefs, Values and Knowledge Systems: leaderboard
Metric: F-score (%; harmonic mean of CO, the share of fully correct answers, and CGA, the correct share of attempted answers, over the 189 questions of this cultural dimension; free-form answers to natively sourced culture-specific short questions asked in their own language, five runs per model, scored by a Gemini-3.1-Pro gold-answer containment judge). Source: arxiv.org. Saturation forecast: Around July 2027. 21 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 71.4 |
| 2 | Claude Opus 4.6 | 58.2 |
| 3 | GPT-5.5 | 55.7 |
| 4 | Seed 2.1 Pro | 54.8 |
| 5 | DeepSeek V4 Pro | 53.1 |
| 6 | GPT-5.4 | 52.1 |
| 7 | Claude Opus 4.7 | 50.2 |
| 8 | Seed 2.0 Pro (High) | 48.5 |
| 9 | DeepSeek V3.2 | 45.6 |
| 10 | Seed 2.0 Pro (Medium) | 45.2 |
| 11 | Qwen 3.5 Plus (Thinking) | 45 |
| 12 | Seed 2.1 Turbo | 44.6 |
| 13 | Kimi K2.6 | 44.4 |
| 14 | GPT-5.2 (High) | 44.3 |
| 15 | Kimi K2.5 | 43.9 |
Interactive version: theaggregate.ai/benchmark?slug=msqa-beliefs-values-and-knowledge-systems · How It Works · Data refreshed daily, snapshot 2026-09-29.