Global PIQA: leaderboard
Multilingual physical commonsense benchmark testing everyday interaction knowledge across many languages and cultural contexts.
Metric: Score (%). Source: mrlbenchmarks.github.io. Source: qwen.ai. Saturation forecast: Estimated already saturated. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.7 Max | 91.4 | #75 |
| 2 | Claude Opus 4.6 (Thinking) | 91.2 | #63 (Claude Opus 4.6) |
| 3 | DeepSeek V4 Pro (Reasoning, Max Effort) | 90.5 | #102 (DeepSeek V4 Pro) |
| 4 | Qwen 3.6 Plus | 89.8 | #120 |
| 5 | GLM-5.1 (Thinking) | 89.5 | #97 (GLM-5.1) |
| 6 | Kimi K2.6 (Thinking) | 89.2 | #104 (Kimi K2.6) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=global-piqa · How It Works · Data refreshed daily, snapshot 2026-10-20.