
DeepSeek Unveils V3.2 Models Claiming Performance Near GPT-5
DeepSeek released its DeepSeek-V3.2 and DeepSeek-V3.2-Speciale models on Monday. The standard V3.2 model aims to balance reasoning capabilities with output length for daily use, with benchmark testing reportedly showing performance on par with OpenAI's GPT-5 and only slightly behind Google’s Gemini-3.0-Pro. The Speciale variant is designed to push the boundaries of reasoning capabilities for open-source models.