China-based AI startup DeepSeek has launched two new models, DeepSeek-V3.2 and DeepSeek-V3.2 Speciale, to challenge the dominance of Google Gemini and OpenAI’s ChatGPT. Following their earlier success with DeepSeek R1 and V3, which matched ChatGPT’s capabilities at a reduced cost, the company faced tough competition from advanced models like GPT-5.1 and Google’s Gemini series. The new V3.2 models are marketed as “reasoning-first” and provide performance comparable to GPT-5 for general tasks. DeepSeek-V3.2-Speciale excels in complex problem-solving, achieving remarkable results in prestigious competitions like the International Mathematical Olympiad (IMO). It outperforms both GPT-5 High and Gemini 3 Pro in math benchmarks and excels in coding challenges. Users can access the V3.2 model directly on the DeepSeek platform, while the Speciale model is currently available via API. These advancements signify DeepSeek’s intent to disrupt the AI chatbot landscape with high-performance models.
Source link
