AI News (2026/1/30): SenseNova-MARS Open-Sourced by SenseTime: Breaking the Ceiling of Multimodal Search and Inference
Executive Summary:
SenseTime has open-sourced the multimodal autonomous reasoning model SenseNova-MARS, offering 8B and 32B versions. The model outperforms Gemini-3-Pro (69.06 points) and GPT-5.2 (67.64 points) with a comprehensive score of 69.74 points in core benchmarks such as MMSearch and HR-MMSearch, becoming the first open-source Agentic VLM to support dynamic visual reasoning and deep integration of image and text search.
SenseTime Open-Sources SenseNova-MARS: Breaking the Ceiling of Multimodal Search and Inference
News Details
On January 30, SenseTime announced the open-sourcing of the multimodal autonomous reasoning model SenseNova-MARS. The model is available in 8B and 32B versions, aimed at advancing the development of multimodal search and inference technologies. SenseNova-MARS excels in multiple core benchmark tests, particularly in MMSearch and HR-MMSearch, where it achieved a comprehensive score of 69.74 points, surpassing Gemini-3-Pro (69.06 points) and GPT-5.2 (67.64 points).
Key Points
Multimodal Autonomous Reasoning: SenseNova-MARS is the first open-source Agentic VLM to support dynamic visual reasoning and deep integration of image and text search. It can handle various modalities of data, including images, text, and video, enabling more comprehensive search and reasoning capabilities.
Performance: In core benchmark tests such as MMSearch and HR-MMSearch, SenseNova-MARS achieved a comprehensive score of 69.74 points, surpassing Gemini-3-Pro (69.06 points) and GPT-5.2 (67.64 points). This indicates its significant advantages in multimodal data processing and reasoning.
Open Access: SenseTime provides both 8B and 32B versions of the SenseNova-MARS model to meet the needs of different developers. The 8B version is suitable for resource-constrained scenarios, while the 32B version offers higher accuracy and stronger reasoning capabilities. Additionally, SenseTime has released detailed documentation and sample code to help developers get started quickly.
AI-ALL In-Depth Commentary
The open-sourcing of SenseNova-MARS marks a significant step forward in multimodal search and inference technologies. As the first Agentic VLM to support dynamic visual reasoning and deep integration of image and text search, it not only outperforms existing mainstream models in terms of performance but also provides developers with flexible options. The release of both 8B and 32B versions ensures that applications of different scales can benefit from it. This initiative is expected to accelerate the application of multimodal AI technologies in real-world scenarios, such as intelligent customer service, content recommendation, and autonomous driving. At the same time, SenseTime's move also challenges existing productivity tools, driving the entire AI ecosystem towards a more open and collaborative direction.
