DeepSeek AI has unveiled DeepSeek-OCR, an innovative optical character recognition system capable of processing over 200,000 pages daily on a single NVIDIA A100 GPU, scaling up to 33 million pages across 20 nodes. The model boasts over 96% precision while compressing text at rates of 9x to 10x, and is designed for high efficiency at various document types, including charts and multilingual text. DeepSeek states this approach could revolutionize text processing in AI applications.
DeepSeek AI has unveiled DeepSeek-OCR, an innovative optical character recognition system capable of processing over 200,000 pages daily on a single NVIDIA A100 GPU, scaling up to 33 million pages across 20 nodes. The model boasts over 96% precision while compressing text at rates of 9x to 10x, and is designed for high efficiency at various document types, including charts and multilingual text. DeepSeek states this approach could revolutionize text processing in AI applications.