DeepSeek Introduces VL2 – an AI Model for Image Analysis

Chinese tech company DeepSeek has unveiled DeepSeek-VL2, an advanced AI model designed to analyze visual content.

Main versions of DeepSeek-VL2:
🔹 DeepSeek-VL2-Tiny – 1 billion parameters
🔹 DeepSeek-VL2-Small – 2.8 billion parameters
🔹 DeepSeek-VL2 – 4.5 billion parameters

VL2 can interpret documents, tables, and diagrams, generating relevant insights. For instance, it can create a recipe based on a food image or explain data from a chart.

At the moment, VL2 is not yet integrated into the DeepSeek chatbot, but the VL2-Small version is available for online testing.
Read 'Zamin' on Telegram!