# PL-MTEB (Polish MTEB)

Tests how well "embedding" models (the part of AI search that turns text into numbers so similar meanings can be found) work on Polish text, on tasks like finding similar sentences, sorting reviews and grouping documents.

Category: polish. Type: leaderboard. Status: active.
Metric: Mean score over all Polish tasks (0-100, from 0-1 scale) (higher is better).
Source: MTEB team (Hugging Face MTEB leaderboard) — https://mteb-leaderboard-backend.hf.space/v1/benchmarks/MTEB(pol, v1)/scores, retrieved 2026-11-17, license: MIT (MTEB project).

| Rank | Model | Mean (Task) | Mean (Task type) | Params (B) |
|---|---|---|---|---|
| 1 | codefuse-ai/F2LLM-v2-14B | 75.13 | 71.96 | 13.99 |
| 2 | codefuse-ai/F2LLM-v2-8B | 74.61 | 71.48 | 7.568 |
| 3 | Qwen/Qwen3-Embedding-8B | 73.54 | 71.66 | 7.567 |
| 5 | Qwen/Qwen3-Embedding-4B | 72.34 | 70.53 | 4.022 |
| 6 | codefuse-ai/F2LLM-v2-4B | 73.42 | 70.28 | 4.022 |
| 7 | Linq-AI-Research/Linq-Embed-Mistral | 71.10 | 69.28 | 7.111 |
| 8 | microsoft/harrier-oss-v1-0.6b | 70.47 | 68.65 | 0.596 |
| 9 | codefuse-ai/F2LLM-v2-1.7B | 72.03 | 68.96 | 1.721 |
| 10 | Salesforce/SFR-Embedding-Mistral | 69.31 | 68.29 | 7.111 |
| 11 | intfloat/multilingual-e5-large-instruct | 69.13 | 68.35 | 0.56 |
| 12 | intfloat/e5-mistral-7b-instruct | 69.94 | 68.66 | 7.111 |
| 13 | Alibaba-NLP/gte-Qwen2-1.5B-instruct | 70.00 | 67.98 | 1.543 |
| 14 | Salesforce/SFR-Embedding-2_R | 68.88 | 67.18 | 7.111 |
| 16 | codefuse-ai/F2LLM-v2-0.6B | 69.63 | 66.72 | 0.596 |
| 17 | manu/bge-m3-custom-fr | 64.74 | 63.88 | 0.568 |
| 17 | microsoft/harrier-oss-v1-270m | 66.93 | 66.04 | 0.268 |
| 21 | OrdalieTech/Solon-embeddings-large-0.1 | 63.49 | 63.01 | 0.56 |
| 22 | codefuse-ai/F2LLM-v2-330M | 66.53 | 64.39 | 0.334 |
| 23 | Qwen/Qwen3-Embedding-0.6B | 65.50 | 64.59 | 0.596 |
| 24 | omarelshehy/arabic-english-sts-matryoshka | 65.07 | 64.30 | 0.56 |
| 26 | sdadas/mmlw-e5-small | 63.41 | 63.35 | 0.118 |
| 28 | HIT-TMG/KaLM-embedding-multilingual-mini-instruct-v1 | 64.57 | 63.70 | 0.494 |
| 30 | hotchpotch/bekko-embedding-v1-a25m | 61.50 | 61.66 | 0.123 |
| 32 | Snowflake/snowflake-arctic-embed-l-v2.0 | 61.31 | 60.96 | 0.568 |
| 35 | deepvk/USER-bge-m3 | 62.46 | 61.27 | 0.359 |

Notes: Benchmark MTEB(pol, v1) from the MTEB leaderboard backend API; source scores are on 0-1 scale and shown here x100 to 2 decimals. Leaderboard Space last modified 2026-11-16. Models with no mean score (not run on all tasks) were skipped. Some models may have trained on part of the tasks (zero-shot share not shown).
