rankings / read-tables
The best models for reading tables
16 models & services · 1 callable here now
Ranked by benchmark score per dollar (quality floor applied). Scores: olmOCR-bench table tests — how accurately a model extracts tables from real documents. Fetched 2026-07-12. Prices are our live per-call rates; ~ marks an estimate until the model is onboarded.
sorted by value · sort by score
| # | model | score | price | params | status |
|---|---|---|---|---|---|
| 1 | deepseek-ai/DeepSeek-OCR best value | 80.2 (#12) | $0.1667 / 1M tokens | 3336.1M | live · use now |
| 2 | datalab-to/chandra-ocr-2 top score | 92.1 (#1) | - | 5295.6M | not hosted |
| 3 | infly/Infinity-Parser2-Pro | 91.2 (#2) | - | 35107.2M | not hosted |
| 4 | rednote-hilab/dots.mocr | 90.7 (#3) | - | 3039.2M | not hosted |
| 5 | tiiuae/Falcon-OCR | 90.3 (#4) | - | 269.9M | not hosted |
| 6 | lightonai/LightOnOCR-2-1B | 89 (#5) | - | 1005.6M | not hosted |
| 7 | rednote-hilab/dots.ocr | 88.3 (#6) | - | 3039.2M | not hosted |
| 8 | datalab-to/chandra | 88 (#7) | - | 8767.1M | not hosted |
| 9 | datalab-to/surya-ocr-2 | 86.6 (#8) | - | 686.2M | not hosted |
| 10 | infly/Infinity-Parser-7B | 85 (#9) | - | 8292.2M | not hosted |
| 11 | opendatalab/MinerU2.5-2509-1.2B | 84.9 (#10) | - | 1156M | not hosted |
| 12 | baidu/Qianfan-OCR | 81.6 (#11) | - | 4741.4M | not hosted |
| 13 | zai-org/GLM-OCR | 77.6 (#13) | - | 1325.3M | not hosted |
| 14 | deepseek-ai/DeepSeek-OCR-2 | 77.4 (#14) | - | 3389.1M | not hosted |
| 15 | lightonai/LightOnOCR-1B-1025 | 76.4 (#15) | - | 1161.2M | not hosted |
| 16 | Mistral OCR | - | $1.00 / 1k pages | - | reference |