In the past month, the global large model market has been shaken by the release of DeepSeek, prompting major companies to be questioned about their views and strategies in response to this emerging force. In the domestic market, six large model startups that have already become unicorns—Zero One Everything, Baichuan Intelligence, LeapStar, Zhipu Huazhang, Dark Side of the Moon, and MiniMax—are facing severe challenges, but their respective strategies show different directions.
Firstly, the founder of Zero One Everything, Kai-Fu Lee, stated earlier that they are no longer pursuing the training of super large models, but are focusing on the development of lightweight models. On February 14, Zero One Everything and the Suzhou High-tech Zone jointly established an industrial large model base, focusing on various fields such as manufacturing, finance, and healthcare, committed to bringing large model technology from the laboratory to practical applications.
Image source note: Image generated by AI, authorized by service provider Midjourney
In just a few days after the release of DeepSeek, Baichuan Intelligence launched the Baichuan-M1-preview, a full-scene inference large model with capabilities in language, vision, and search, and is focusing on transitioning to the healthcare industry with the introduction of the "AI Pediatrician" project, demonstrating its rapid implementation in practical applications.
LeapStar chose to launch two new models on the same day as the release of DeepSeek to enhance its market competitiveness. Additionally, its application "Yuewen" introduced new features aimed at helping users develop applications without code, further enhancing the user experience.
Tang Jie, the founder of Zhipu Huazhang, emphasized at an international summit that future large language models will become central to daily life and work, completing complex tasks through autonomous learning and collaboration with humans, showcasing his vision for the future.
Shortly after the release of DeepSeek, Dark Side of the Moon launched the Kimi k1.5 multimodal thinking model to enhance the model's contextual understanding and strategy optimization, indicating its ongoing efforts in technological innovation.
Although these six startups did not directly respond to DeepSeek, their actual actions and product innovations are actively addressing this impact and demonstrating their ability to adapt to market changes.
