PL-MTEB: leaderboard
Polish Massive Text Embedding Benchmark leaderboard evaluating embedding models across Polish classification, clustering, pair classification, semantic textual similarity, and retrieval tasks.
Metric: Mean Main Score (%). Source: huggingface.co. 22 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | gte-Qwen2-7B-instruct | 65.86 |
| 2 | bge-multilingual-gemma2 | 64.72 |
| 3 | gte-Qwen2-1.5B-instruct | 62.85 |
| 4 | mmlw-roberta-large | 61.87 |
| 5 | mmlw-roberta-base | 59.79 |
| 6 | mmlw-e5-large | 59.74 |
| 7 | KartonBERT-USE-base-v1 | 59.04 |
| 8 | mmlw-e5-base | 58.54 |
| 9 | multilingual-e5-large | 58.5 |
| 10 | bge-m3 | 57.19 |
Interactive version: theaggregate.ai/benchmark?slug=pl-mteb · How It Works · Data refreshed daily, snapshot 2026-09-05.