UrduMMLU (Urdu Prompt) - STEM: leaderboard
The STEM domain of UrduMMLU (2026), an Urdu MMLU-style benchmark from native question banks and Pakistani examinations: exact-match accuracy in percent on biology, chemistry, computer science, general science, mathematics and physics questions, with an Urdu instruction prompt, zero-shot. Higher is better.
Metric: STEM (5,113 questions: biology, chemistry, computer science, general science, mathematics, physics) exact-match accuracy (%) on UrduMMLU's Urdu multiple-choice questions from Pakistani MCQ banks and public examinations, zero-shot, answer key parsed from the generation (unparsable, malformed or error outputs count as wrong); temperature 0 where available, 4,096 output tokens, Urdu instruction prompt. Source: arxiv.org. Saturation forecast: Estimated already saturated. 30 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.5 Flash | 97.75 | #101 |
| 2 | GPT-5.4 | 97.4 | #123 |
| 3 | DeepSeek V4 Flash | 97.36 | #258 |
| 4 | Gemini 3.1 Flash Lite | 97.09 | #364 |
| 5 | Claude Sonnet 4.6 | 96.26 | #161 |
| 6 | Qwen 3.6 35B A3B | 96.15 | #462 |
| 7 | Gemma 4 31B (IT) | 93.86 | #342 |
| 8 | Llama 4 Maverick Instruct | 92.24 | #1133 |
| 9 | Claude Haiku 4.5 | 91.92 | #568 |
| 10 | Qwen 3.6 27B | 91.12 | #341 |
| 11 | GPT-5.4 Mini | 88.25 | #389 |
| 12 | Gemma 4 26B A4B (IT) | 87.19 | #549 |
| 13 | Llama 4 Scout Instruct | 85.39 | #1543 |
| 14 | Llama 3.3 70B Instruct | 78.39 | #1357 |
| 15 | Qwen 3 8B | 74.05 | #1830 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=urdummlu-urdu-prompt-stem · How It Works · Data refreshed daily, snapshot 2026-11-13.