Loading page…
Preferred languages
More languages supported by Google Translate
Open the selector to load the current language catalog.
Translations are provided by Google Translate.
Loading page…
Preferred languages
More languages supported by Google Translate
Open the selector to load the current language catalog.
Translations are provided by Google Translate.
Multimodal & Grounded benchmark
Every model the catalog carries a published OCRBench V2 value for, ranked by that value.
A native OCR benchmark for reading text from images across multilingual scripts, low-quality scans, handwriting, structured layouts, charts, and screenshots. 9 models in the ModelChoice catalog have a published value.
Each point is a model with a published OCRBench V2 value and complete token prices, on the 75% input / 25% output blend. Frontier models are spaced evenly along the cost axis.
9 models with a published OCRBench V2 value, ordered by that value, highest first. Models the source has not scored on this benchmark are not listed — they are unmeasured, not last.
| Rank | Model | Provider | Accuracy |
|---|---|---|---|
| 1 | Alibaba | 74.2 | |
| 2 | INInterfaze Beta | Interfaze | 70.7 |
| 2 | Alibaba | 70.7 | |
What the benchmark measures and how its value is published, from the source’s own record.
OCRBench V2 is a multimodal & grounded benchmark published by OCRBench v2: An Improved Benchmark for Evaluating Large Multimodal Models on Visual Text Localization and Reasoning. A native OCR benchmark for reading text from images across multilingual scripts, low-quality scans, handwriting, structured layouts, charts, and screenshots. Its evaluation set is described by the source as image OCR tasks. The source rates its difficulty as native visual text understanding.
Each model's result is published by OCRBench v2: An Improved Benchmark for Evaluating Large Multimodal Models on Visual Text Localization and Reasoning as accuracy. ModelChoice prints that value to 1 decimal place, exactly as published, and never converts or rescales it. This page ranks the 9 models that carry a published value, highest value first; a model the source never scored is left out rather than ranked at zero. The source does not state whether a higher value is the better result, so neither does this page.
All leaderboards · Multimodal & Grounded capability leaderboard
| 4 |
| Alibaba |
| 67.1 |
| 5 | Z.AI | 65.1 |
| 6 | Z.AI | 63.5 |
| 6 | LiquidAI | 63.5 |
| 8 | PMTernary Bonsai 2 27B | Prism ML | 56.9 |
| 9 | LiquidAI | 47.5 |
Evidence key: Observed
Rows are ordered by the value OCRBench v2: An Improved Benchmark for Evaluating Large Multimodal Models on Visual Text Localization and Reasoning published, highest first. The source does not state whether a higher value is the better result, so this page does not either: for a benchmark that measures a rate of failure, read the table from the bottom.