MSQA - Thai: leaderboard
Metric: F-score (%; harmonic mean of CO, the share of fully correct answers, and CGA, the correct share of attempted answers, over the 95 Thai-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 May 2027. 21 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 71 |
| 2 | GPT-5.5 | 58.5 |
| 3 | Claude Opus 4.6 | 51 |
| 4 | DeepSeek V4 Pro | 49.5 |
| 5 | GPT-5.4 | 48.6 |
| 6 | Seed 2.1 Pro | 45.9 |
| 7 | GPT-5.2 (High) | 45.1 |
| 8 | Gemini 2.5 Flash | 40.2 |
| 9 | Qwen 3.5 Plus (Thinking) | 39.3 |
| 10 | DeepSeek V3.2 | 34.8 |
| 11 | GLM-5 | 34.8 |
| 12 | Seed 2.0 Lite | 34.3 |
| 13 | Seed 2.0 Pro (High) | 30.9 |
| 14 | Kimi K2.6 | 28.3 |
| 15 | Kimi K2.5 | 27.9 |
Interactive version: theaggregate.ai/benchmark?slug=msqa-thai · How It Works · Data refreshed daily, snapshot 2026-09-29.