DROP
22 models
Top 5 Models Performance
| Tencent: youtu-llm-2b | ######################################## | 86.7 |
| Xiaomi: MiMo-V2.5 Pro | ######################################## | 86.3 |
| Tencent: hy3-preview-base | ####################################### | 85.5 |
| Google: palm | ####################################### | 85 |
| Qwen: Qwen3 4B | ###################################### | 82.9 |
| Rank | Model | Score | |
|---|---|---|---|
| 🇨🇳 | 1 | Tencent: youtu-llm-2b腾讯控股有限公司 | 86.7 |
| 🇨🇳 | 2 | Xiaomi: MiMo-V2.5 Pro小米集团 | 86.3 |
| 🇨🇳 | 3 | Tencent: hy3-preview-base腾讯控股有限公司 | 85.5 |
| 🇺🇸 | 4 | Google: palm | 85 |
| 🇨🇳 | 5 | Qwen: Qwen3 4B通义千问 | 82.9 |
| ❓ | 6 | sapientinc: hrm-text-1b | 82.2 |
| 🇺🇸 | 7 | OpenAI: gpt-4 | 80 |
| 🇺🇸 | 8 | Google: Gemma 3 27B Pretrained | 77.2 |
| 🇨🇳 | 9 | Qwen: qwen3-1.7b通义千问 | 72.5 |
| 🇺🇸 | 10 | Google: gemma-3-12b-pt | 72.2 |
| 🇺🇸 | 11 | Hugging Face Smol Models Research: smollm3-3b | 72 |
| 🇺🇸 | 12 | OpenAI: GPT-3.5 Turbo | 70.2 |
| 🇷🇺 | 13 | Yandex: gpt-5-lite-pretrainЯндекс | 67.8 |
| 🇷🇺 | 14 | Yandex: YandexGPT Lite 5Яндекс | 65.1 |
| 🇺🇸 | 15 | OpenAI: GPT-3.5 | 64.1 |
| 🇺🇸 | 16 | Google: gemma-3-4b-pt | 60.1 |
| 🇷🇺 | 17 | Sber: gigachat3.5-432b-a28b-baseСбер | 59.55 |
| 🇺🇸 | 18 | Meta Llama: Llama 3.2 3B | 45.2 |
| 🇺🇸 | 19 | Google: gemma-3-1b-pt | 42.4 |
| 🇨🇳 | 20 | DeepSeek: deepseek-r1-distill-qwen-1.5b深度求索 | 41.3 |
| 🇨🇳 | 21 | Qwen: qwen3.5-2b-base通义千问 | 30.8 |
| 🇺🇸 | 22 | Arcee AI: trinity-nano-base | 4.5 |