SenseTime’s SenseNova Achieves Two Top Titles Again

Defending the Dual Titles

SenseTime's "SenseNova V6" (SenseNova-V6 Reasoner) has secured "dual firsts" once again, ranking joint first in China for general language capabilities and achieving the top global spot for multimodal reasoning, according to the May 2025 Chinese Large Model Benchmark Evaluation Report released by SuperCLUE, a leading large model evaluation institution, on May 28. Scoring an outstanding 62.96 points, it tied with Doubao-1.5-thinking-pro for the top domestic position in China, claiming the gold medal, and also earned the highest domestic score in the Agent testing task.

Meanwhile, according to data from Hugging Face, the world's largest AI open-source community, SenseTime's "SenseNova V6" (SenseNova-V6 Pro) ranked first globally in the multimodal large model evaluation list of Open Compass this month, surpassing Gemini 2.5 Pro with an overall score of 80.4, making it the world's most powerful multimodal large model.

This marks the second time in six months that SenseTime's SenseNova large model has won "dual firsts"! Earlier this year, SenseNova achieved “dual firsts” in the 2024 Annual Chinese Large Model Benchmark Evaluation Report by SuperCLUE and the multimodal evaluation of OpenCompass using the same model, drawing intense attention and leading the industry trend toward unifying large language models and multimodal large models.

This successful defense of the "dual firsts" – winning top spots in both general language and multimodal authoritative rankings – signifies that SenseTime's "SenseNova V6" has become one of the world's most advanced large models, continuously pushing the boundaries of AI capabilities in multimodality.

Through technological innovations such as multimodal long chain-of-thought training, global memory, and reinforcement learning, "SenseNova V6" can handle various tasks, including text and multimodal processing, with a single model. It boasts significant advantages in reasoning, mathematics, and coding, enabling deep understanding of human intentions and processing complex information and problems to provide robust support for real-world business applications. Additionally, its multimodal training efficiency is comparable to language training, reaching the industry's best level, while its inference costs are the lowest in the industry.

With its strong balance of high performance and cost efficiency, "SenseNova V6" has gained widespread industry support since its launch in April this year. It has partnered with multiple enterprises in fields such as embodied intelligence, smart education, and smart culture-tourism to jointly promote multi-industry and multi-scenario implementation.

Adhering to the trinity development strategy of "AI infrastructure - large model - applications," SenseTime continues to enhance multimodal and deep-thinking capabilities while reducing training and inference costs through the joint optimization of AI infrastructure systems and model algorithms, driving the rapid development of generative AI and its integration into "daily life for all."