MSQA - Indonesian: leaderboard
Metric: F-score (%; harmonic mean of CO, the share of fully correct answers, and CGA, the correct share of attempted answers, over the 81 Indonesian-language questions; 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 January 2028. 21 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 62.9 |
| 2 | Claude Opus 4.7 | 61.3 |
| 3 | Claude Opus 4.6 | 48.5 |
| 4 | GPT-5.5 | 45.4 |
| 5 | Seed 2.1 Pro | 45.2 |
| 6 | GPT-5.4 | 45 |
| 7 | DeepSeek V4 Pro | 43.1 |
| 8 | DeepSeek V3.2 | 42.8 |
| 9 | Seed 2.0 Pro (High) | 42.6 |
| 10 | Gemini 2.5 Flash | 40.6 |
| 11 | Seed 2.1 Turbo | 40 |
| 12 | Qwen 3.5 Plus (Thinking) | 38.9 |
| 13 | GPT-5.2 (High) | 37.3 |
| 14 | Seed 2.0 Pro (Medium) | 37.3 |
| 15 | Kimi K2.6 | 36.7 |
Interactive version: theaggregate.ai/benchmark?slug=msqa-indonesian · How It Works · Data refreshed daily, snapshot 2026-09-29.