GSM-Plus-BN - Critical Thinking: leaderboard
GSM-Plus-BN (2026) is a human-verified Bangla translation of the GSM-Plus perturbation benchmark for grade-school math. Accuracy in percent on the critical thinking variants (essential information removed, so the model must say the question cannot be answered) with standard prompting; higher is better.
Metric: Accuracy (%; critical thinking perturbation, essential information removed, so the model must say the question cannot be answered; Bangla grade-school math word problems translated from GSM-Plus and verified by human translators; standard prompting (answer directly); Groq API). Source: arxiv.org. Saturation forecast: Around February 2027. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-OSS-120B | 28.6 | #700 |
| 2 | GPT-OSS-20B | 25.2 | #1147 |
| 3 | Llama 4 Scout Instruct | 21.4 | #1543 |
| 4 | Qwen 3 32B | 6.6 | #960 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=gsm-plus-bn-critical-thinking · How It Works · Data refreshed daily, snapshot 2026-11-13.