MSQA - Chinese: leaderboard
Metric: F-score (%; harmonic mean of CO, the share of fully correct answers, and CGA, the correct share of attempted answers, over the 150 Chinese-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 2029. 21 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 56.8 |
| 2 | Seed 2.0 Pro (High) | 55.7 |
| 3 | DeepSeek V4 Pro | 50.9 |
| 4 | Seed 2.1 Pro | 50.7 |
| 5 | Seed 2.0 Pro (Medium) | 50.2 |
| 6 | Seed 2.1 Turbo | 48.7 |
| 7 | DeepSeek V3.2 | 45.3 |
| 8 | GPT-5.5 | 44.5 |
| 9 | Claude Opus 4.6 | 43.9 |
| 10 | GLM-5 | 43.5 |
| 11 | Seed 2.0 Lite | 42.2 |
| 12 | Kimi K2.5 | 41.5 |
| 13 | Kimi K2.6 | 40.9 |
| 14 | GPT-5.4 | 38.3 |
| 15 | Qwen 3.5 Plus (Thinking) | 34.9 |
Interactive version: theaggregate.ai/benchmark?slug=msqa-chinese · How It Works · Data refreshed daily, snapshot 2026-09-29.