Leaderboard

davanstrien/bpl-ocr-bench-results

Rankings are computed using Bradley-Terry MLE from pairwise comparisons judged by a vision-language model. The judge sees the original document image alongside two anonymised OCR outputs and picks the more faithful transcription. Browse the comparisons to see the evidence — and vote yourself to build a Human ELO column. Human votes are stored locally for this session only and will reset when the server restarts.

#	Model	Params	Judge ELO	95% CI	Wins	Losses	Ties	Win%
1	LightOnOCR-2-1B	1B	1559	1497–1630	39	25	0	61%
2	GLM-OCR	0.9B	1535	1471–1591	48	35	1	57%
3	dots.ocr	1.7B	1453	1385–1515	26	37	0	41%
4	DeepSeek-OCR	4B	1452	1388–1514	33	49	1	40%

ELO vs Parameter Count

Smaller models can win on the right documents. Error bars show 95% confidence intervals.