In 2025, the global AI landscape was dominated by Google’s Gemini series and OpenAI’s GPT-5 advancements. However, the Chinese startup DeepSeek has made a notable comeback, unveiling its powerful new models, V3.2 and V3.2 Speciale, aimed at rivaling these established giants. DeepSeek initially gained attention in early 2025 with its cost-effective models but faded from view as competitors flourished. Its latest offerings boast reasoning capabilities comparable to GPT-5 and Google’s Gemini 3 Pro, specifically designed for agentic workflows. Notably, V3.2 Speciale achieved impressive results in elite math benchmarks, outperforming top models in pure math and coding logic tests. While it still lags behind on certain benchmarks like Humanity’s Last Exam, DeepSeek’s resurgence signals a competitive shift in AI, challenging the dominance of Western firms. As it gradually releases access to users, the AI race has intensified, promising exciting developments ahead.
Source link
