DeepSeek challenges Google and OpenAI with a new AI model.

(CLO) DeepSeek has just announced its most powerful variant, DeepSeek-V3.2-Speciale, which is said to rival Google DeepMind's new Gemini 3 Pro in some tasks.

In an announcement on Monday, Hangzhou-based DeepSeek stated that its V3.2-Speciale is on par with Google's Gemini 3 Pro (released two weeks ago) in terms of inference capabilities.

DeepSeek trains AI models at surprisingly low costs.

Meanwhile, the base model of V3.2-Speciale – which is V3.2 released on the same day – has performance comparable to OpenAI's GPT-5 (released in August).

DeepSeek announced that V3.2-Speciale achieved a gold medal in the International Mathematical Olympiad, an achievement previously only attained by unpublished internal models from OpenAI and Google DeepMind.

DeepSeek has released the open-source code for its V3.2 model on the Hugging Face developer platform. However, according to the announcement, V3.2-Speciale can only be accessed through an application programming interface (API).

DeepSeek achieved this feat despite the Chinese company having limited access to advanced semiconductor chips.

Source: https://congluan.vn/deepseek-thach-thuc-google-va-openai-voi-moi-hinh-ai-moi-10320085.html