Konkur 1404 - Mathematics: leaderboard
The Mathematics stream of Konkur 1404 — Persian-language multiple-choice questions from the mathematics exam of Iran's 1404 (2025) national university entrance exam; text-only accuracy (%).
Metric: Accuracy (%, text-only). Source: huggingface.co. Saturation forecast: Around January 2027. 23 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-OSS-120B | 68.7 | #341 |
| 2 | gpt-oss-safeguard-20B | 66.96 | #314 |
| 3 | Qwen 3 VL 235B A22B Instruct | 60 | #273 |
| 4 | Kimi K2 0905 | 52.17 | #288 |
| 5 | Qwen 3 Next 80B A3B Instruct | 47.83 | #368 |
| 6 | Qwen 3 VL 30B A3B Instruct | 46.96 | #384 |
| 7 | DeepSeek V3.2 | 45.22 | #205 |
| 8 | Gemma 3 27B (IT) | 44.35 | #523 |
| 9 | Devstral Small 2 | 44.35 | #495 |
| 10 | Llama 3.1 70B Instruct | 42.61 | #561 |
| 11 | MiMo-V2-Flash | 40 | #372 |
| 12 | Ministral 3 14B | 35.65 | #650 |
| 13 | Gemma 3n E4B (IT) | 34.78 | #749 |
| 14 | Ministral 8B | 33.91 | #1040 |
| 15 | Mixtral 8x22B Instruct | 33.04 | #728 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=konkur-1404-mathematics · How It Works · Data refreshed daily, snapshot 2026-10-20.