Open PL LLM - KLEJ NER (multiple choice, 5-shot): leaderboard
Named-entity category classification of Polish text fragments from the KLEJ benchmark (NKJP-NER). Open PL LLM Leaderboard (SpeakLeash), 5-shot, multiple choice variant: the options are ranked by log-likelihood; accuracy (%), higher is better. Runs not available on the Hugging Face Hub are excluded.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Around February 2027. 321 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Mistral Large 2 (Nov) Instruct (2411) | 65.16 | #480 |
| 2 | internlm2-chat-20B | 64.67 | #988 |
| 3 | Qwen 3 32B | 64.29 | #449 |
| 4 | Mistral Large 2 (Jul) | 63.51 | #490 |
| 5 | Qwen 2.5 32B Instruct | 63.07 | #509 |
| 6 | QwQ 32B-Preview | 62.68 | #671 |
| 7 | Llama 3.1 405B Instruct FP8 | 62.63 | #400 |
| 8 | Qwen 2.5 32B | 62.54 | #544 |
| 9 | Qwen 2 72B Instruct | 62.2 | #579 |
| 10 | Phi-4 | 61.27 | #707 |
| 11 | Qwen 2.5 72B Instruct | 61.03 | #469 |
| 12 | Qwen 1.5 32B | 60.54 | #794 |
| 13 | Qwen 3 14B | 60.3 | #542 |
| 14 | Llama 3 70B Instruct | 60.3 | #640 |
| 15 | Qwen 1.5 72B | 59.77 | #752 |
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-klej-ner-multiple-choice-5-shot · How It Works · Data refreshed daily, snapshot 2026-11-05.