China strikes back in AI wars! DeepSeek unleashes V3.2 & V3.2 Speciale -reasoning-first models claiming to match OpenAI’s GPT-5 & rival Google’s Gemini 3.0-Pro. Built for agents with integrated tool-use, massive 1,800-sim environments, & 85K+ instructions. Powered by DeepSeek Sparse Attention (DSA) for faster, cheaper long-prompts. Gold-medal wins at 2025 Math & Informatics Olympiads push into AGI territory. After months quiet, Hangzhou’s rising star challenges US dominance amid Qwen & Kimi buzz. Will DeepSeek dethrone ChatGPT?
Credit to : Mint
Please support our Sponsors here :