Shanghai, April 11 — In a bold move to redefine its place in the global AI race, Chinese artificial intelligence powerhouse SenseTime has unveiled its next-generation multimodal models, SenseNova V6 and V6 Reasoner , which it claims outperform industry heavyweights like OpenAI and Google in advanced reasoning tasks.
Announced on Thursday, the new models represent a strategic shift in AI development—moving beyond traditional text-based large language models (LLMs) to multimodal AI , capable of integrating and processing diverse inputs like text, images, audio, and video.
According to SenseTime CEO and Chairman Xu Li , SenseNova V6 , with 600 billion parameters , is not only the most powerful multimodal model developed in China but also the most cost-efficient for inference tasks globally. Xu cited independent benchmarking data from TableBench , showing V6 outperforming OpenAI’s GPT-4o in key areas such as fact-checking, numerical reasoning, data analysis, and visualization .
Meanwhile, the V6 Reasoner has reportedly bested OpenAI’s o1 and Google’s Gemini 2.0 Flash Thinking in multimodal reasoning capabilities, solidifying SenseTime’s position at the cutting edge of AI development.
Xu noted that the conventional approach of scaling models using internet-sourced text data has reached its limit. “We’ve nearly exhausted all available high-quality textual data,” he stated. SenseTime’s answer? Feed the models with multimodal data , leading to unexpected improvements in textual understanding as well.
The company predicts that 2025 will mark the true rise of multimodal AI models , driven by advancements in reinforcement learning and real-world interaction . However, unlike some global rivals, SenseTime remains cautious about open-sourcing its models, citing commercial incentives. “Open source needs a purpose,” Xu remarked, though he left the door open for future possibilities if meaningful industry value is identified.
SenseTime’s push toward monetization is already showing results. In 2024, the firm’s generative AI business accounted for 63.7% of total revenue , up from 34.8% in 2023—surpassing its long-dominant computer vision segment . Overall revenue grew 11% year-on-year to 3.8 billion yuan (US$518 million) , while net losses shrank from 6.5 billion yuan to 4.3 billion yuan , a result of tighter expense management.
To demonstrate real-world applications, SenseTime introduced AI-powered chatbots , office tools , and code-generation platforms at the event. The company also revealed a new partnership with Fourier Intelligence , a Shanghai-based robotics startup. The collaboration aims to integrate V6 models into humanoid robots , enabling them to understand and respond to the world through multimodal perception.
Founded in Hong Kong in 2014 and publicly listed since 2021 , SenseTime is now staking its future on its ability to commercialize cutting-edge AI , with profitability in 2025 as the ultimate goal.
