SenseTime's Lin Dahua: AI Models Need to Balance Performance and Cost, Drawing Inspiration from DeepSeek and Integrating Seedance Capabilities

SenseTime co-founder and chief scientist Lin Dahua recently stated that the company drew inspiration from DeepSeek: even with limited funds and technical resources, high-performance AI models can still be launched.

Lin Dahua mentioned that OpenAI's ChatGPT Images 2.0 can generate exquisite images, but SenseTime's Rixin U1 costs only one-tenth of the former. He noted that while there is a gap with international cutting-edge models like GPT Image 2 and Gemini Nano Banana, their own models are cheaper and more efficient, and most tasks do not require top-tier models.

In response to the competitive pressure from ByteDance's AI video model Seedance, SenseTime has integrated some of its capabilities into the short video tool Seko, combining background generation with its own audio functions. Lin Dahua emphasized that SenseTime aims to create differentiation by integrating large models, applications, and infrastructure, improving service quality while reducing per-use costs; enterprise customers value differentiated value more, and price wars are unsustainable.