Open PL LLM - Belebele (generative, 5-shot): leaderboard
Belebele four-option reading comprehension on FLORES-200 passages in Polish. Open PL LLM Leaderboard (SpeakLeash), 5-shot, generative variant: the answer is extracted from generated text; accuracy (%), higher is better. Runs not available on the Hugging Face Hub are excluded.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 319 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Mistral Large 2 (Nov) Instruct (2411) | 93.78 | #480 |
| 2 | DeepSeek V3 (0324) | 93.67 | #349 |
| 3 | QwQ 32B-Preview | 93.44 | #671 |
| 4 | Llama 3.1 405B Instruct FP8 | 93.22 | #400 |
| 5 | Mistral Large 2 (Jul) | 93 | #490 |
| 6 | Llama 3.3 70B Instruct | 92.78 | #541 |
| 7 | Qwen 2.5 32B | 92.56 | #544 |
| 8 | Qwen 2.5 32B Instruct | 92.44 | #509 |
| 9 | Llama 3.1 70B Instruct | 92 | #567 |
| 10 | QwQ-32B | 92 | #422 |
| 11 | Mistral Small 3 | 92 | #630 |
| 12 | Qwen 2.5 72B | 92 | #512 |
| 13 | Llama 3.1 Nemotron 70B Instruct | 92 | #563 |
| 14 | Qwen 2.5 72B Instruct | 91.89 | #469 |
| 15 | Qwen 3 32B | 91.89 | #449 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=open-pl-llm-belebele-generative-5-shot · How It Works · Data refreshed daily, snapshot 2026-11-05.