DeepSeek V4 Flash is an efficiency-optimized Mixture-of-Experts model from DeepSeek with 284B total parameters and 13B activated parameters, supporting a 1M-token context window. It is designed for fast inference and high-throughput workloads, while maintaining strong reasoning and coding performance.
DeepSeek-OCR 2 is a model released by DeepSeek on January 27, 2026, using the innovative DeepEncoder V2 method, which allows AI to dynamically rearrange parts of an image based on its meaning, rather than just mechanically scanning from left to right. While maintaining high data compression efficiency, the model has achieved significant breakthroughs in multiple benchmarks and production metrics. The model can cover complex document pages with only 256 to 1120 vision tokens, achieving an overall score of 91.09% in the OmniDocBench v1.5 evaluation
deepseek-r2 coming soon
DeepSeek-OCR is an optical character recognition model for extracting text from images and documents. It processes scanned pages, photos, and UI screenshots to produce transcriptions with layout cues such as line breaks. Common uses include document digitization, invoice and receipt intake, search indexing, and enabling RPA pipelines. Technical highlights include image-to-text processing, support for scanned and photographed content, and structured text output for downstream parsing.
A 671B parameter Mixture of Experts text generation model, merged from DeepSeek-AI's R1-0528, R1, and V3-0324, supporting up to 60k tokens of context.