LightOn launches LightOnOCR-3: New document intelligence models

LightOn, a company specializing in artificial intelligence, announces the launch of LightOnOCR-3, a new range of AI models for document intelligence. These models stand out for their high performance on major international benchmarks, often surpassing American and Chinese competitors. In addition to text recognition, LightOnOCR-3 excels in extracting graphic data and describing images, ensuring comprehensive and optimized processing of complex documents.

The available versions of LightOnOCR-3, with 0.8 billion and 4 billion parameters, are offered under the Apache 2.0 license. They meet the diverse needs of businesses, offering twice the processing speed of previous models while being more cost-effective. By combining text transcription and image processing, these models greatly facilitate administrative processes, technical research, and financial analysis.