LaBSE: benchmark results
Provider: Google. Released 2020-08-01. Access: Open.
Aggregate Rating 1540 ± 32, rank #1836 of 10916 rated models, from 140 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| EuroEval Catalan NLU - WikiANN CA | 83.21 | Micro F1 excluding MISC (%) | 100 |
| EuroEval Croatian NLU - WikiANN HR | 82.1 | Micro F1 excluding MISC (%) | 100 |
| EuroEval Czech NLU - PONER | 72.13 | Micro F1 excluding MISC (%) | 100 |
| EuroEval Lithuanian NLU - WikiANN LT | 79.94 | Micro F1 excluding MISC (%) | 100 |
| EuroEval Portuguese NLU - HAREM | 78.12 | Micro F1 excluding MISC (%) | 100 |
| EuroEval Serbian NLU - UNER SR | 92.01 | Micro F1 excluding MISC (%) | 100 |
| SkMTEB - Bitext Mining | 97.48 | Mean F1 (%) over the six bitext-mining datasets; higher is b | 100 |
| EuroEval Spanish NLU - CoNLL ES | 85.12 | Micro F1 excluding MISC (%) | 99.8 |
| EuroEval Estonian NLU - EstNER | 69.11 | Micro F1 excluding MISC (%) | 99.6 |
| EuroEval Latvian NLU - Fullstack NER LV | 80.08 | Micro F1 excluding MISC (%) | 99.6 |
| EuroEval Albanian NLU - WikiANN SQ | 86.4 | Micro F1 excluding MISC (%) | 99.5 |
| EuroEval Bosnian NLU - WikiANN BS | 84.73 | Micro F1 excluding MISC (%) | 99.5 |
Interactive version: theaggregate.ai/model?slug=labse · How It Works · Data refreshed daily, snapshot 2026-11-08.