DeepSeek Strikes Back: China’s Ambitious New AI Models Aim Directly at Gemini and ChatGPT

DeepSeek, a Chinese AI startup, has re-entered the global AI race with its new AI models, V32 and V3.2 Speciale, claiming to offer GPT-5-level performance and reasoning capabilities comparable to Google’s Gemini line.

After a period of silence, DeepSeek aims to reclaim its position as a top contender in the AI landscape, challenging major players like Google and OpenAI with its ambitious new releases

The V32 Speciale model has achieved gold-medal-level performance in prestigious competitions like the 2025 International Mathematical Olympiad and the International Olympiad in Informatics, highlighting its mathematical prowess.

DeepSeek claims that its models outperform leading AI systems from OpenAI and Google on math and coding benchmarks, although its performance on certain complex tests like Humanity’s Last Exam still lags behind Google’s Gemini 3 Pro

The V32 model is now available on the DeepSeek website and app, while the V3.2 Speciale is accessible only through an API, making it a limited release similar to how Google and OpenAI manage their high-power variants.

The comeback of DeepSeek signifies a renewed competitive edge in the realm of reasoning-first agentic AI, posing a challenge to both American and Chinese AI leaders

By achieving Olympiad-level success on a startup budget, DeepSeek strengthens China's position in math-intensive AI research, showcasing its potential to innovate cost-efficiently

The re-emergence of DeepSeek pressures major AI labs like OpenAI and Google to continue pushing the boundaries of innovation, as the competition becomes increasingly fierce

DeepSeek's return marks a pivotal moment in the AI industry, suggesting that the competitive dynamics are evolving with more players capable of challenging established giants

As DeepSeek rolls out its models to more users, the true test will be how these AI systems perform on a global scale, adding intrigue and intensity to the ongoing AI race