SenseNova-SI - SenseTime's Open-Source Spatial Intelligence Large Model

What is SenseNova-SI

SenseNova-SI is SenseTime's open-source spatial intelligence large model, focused on enhancing spatial intelligence. Through training on large-scale, high-quality spatial data, the model significantly enhances its capabilities in core dimensions such as spatial measurement, relationship understanding, and perspective transformation. In multiple authoritative benchmark tests, SenseNova-SI surpasses open-source models of the same scale, leading GPT-5 and other top closed-source models. The model provides detailed installation and usage guides to help developers get started quickly, driving the development of embodied intelligence and world models, and laying the foundation for AI to understand the three-dimensional world.

SenseTime's latest open-source spatial intelligence model, SenseNova-SI-1.3, ranked first in the EASI-8 comprehensive evaluation, surpassingGemini-3-Pro, with outstanding performance especially on high-difficulty tasks such as perspective transformation and spatial reasoning. By constructing hierarchical training data and verifying the "spatial intelligence scale effect," the model proves that high-quality data is more effective than simply increasing model parameters in improving spatial understanding capabilities, providing a powerful open-source foundation for applications such as embodied intelligence.

Main Features of SenseNova-SI

• Spatial Measurement and Estimation: The model can accurately quantify and estimate the size, distance, etc. of objects.
• Spatial Relationship Understanding: The model can understand the relative position, direction, and spatial layout between objects.
• Perspective Transformation: Supports handling information changes when observing the same scene from different perspectives, and inferring the impact of perspective changes.
• Spatial Reconstruction and Deformation: Understands the three-dimensional structure of objects, and maintains spatial cognition after deformation or reconstruction.
• Spatial Reasoning: Performs logical reasoning based on spatial information, such as determining the direction of object movement or changes in spatial layout.
• Multimodal Fusion: Combines multimodal data such as images and text to enhance understanding of complex spatial scenes.

Technical Principles of SenseNova-SI

• Scale Effect: Through training on large-scale, high-quality spatial data, the "scale effect" is verified, i.e., the increase in data volume can significantly enhance the model's spatial cognition ability, which is the core reason for the performance leap of SenseNova-SI.
• Systematic Training Method: SenseTime proposed a spatial capability classification system, based on which the data scale was expanded, and a systematic training method was adopted to achieve consistent improvement of the model across multiple spatial intelligence dimensions.
• Multimodal Fusion Architecture: Based on InternVL and other foundational architectures, SenseNova-SI can effectively fuse image and text information to enhance understanding of complex scenes.