OCR training runs
Reading scanned Sanskrit books: an open vision-language model fine-tuned for page OCR, and the bulk OCR jobs that turn scans into corpus text.
Headline score: Median letter error rate · lower is better The share of letters read wrong on 54 held-out book pages, median over pages: 0.0096 means 0.96% of letters.Letter error rate: the share of letters an OCR model reads wrong, as a fraction (0.0096 means 0.96% of letters). Lower is better.
Live now
Progress
Lower is better on the headline.
Best Median letter error rate over time
Each dot is a scored run, placed at the day it ended; the line is the best score so far. Numbered lines are this track's milestones, listed below.
Milestones
- 5 Oct 2026 · OCR-VLM pilot: first OCR run
- 6 Oct 2026 · OCR-VLM v1 beats Google Vision on Sanskrit pages
Model size and Median letter error rate
Parameters (log scale) against Median letter error rate: this track's runs. Hover a point for its name.
All OCR runs
Published · 4 runs
| Data | |||||||
|---|---|---|---|---|---|---|---|
| OCR pilot test ocr_pilot100k_test |
6 Oct 2026 | — | — | 0.57 | $0.52 | — | — |
| OCR-VLM v1 ocrvlm_v1_q35 |
5 Oct 2026 | — | OCR page labels | 7.33 | $9.53 | Winner | 0.0096 |
| OCR-VLM pilot ocrvlm_pilot_q35 |
4 Oct 2026 | — | OCR page labels | 1.37 | $1.80 | Keep | — |
| OCR pilot 100k Running ocr_pilot100k |
— | — | — | — | $21.63 | — | — |
No runs match these filters. Clear the search or pick “All”.
Cost “—” means no cloud bill (our own GPU at home) or a cost that was not recorded; “Free” is free cloud compute. GPU-h is GPU hours.