SenseTime has introduced the SenseNova U1.5-Lite-Preview model, which has been opened to the community. This lightweight unified multimodal model continues to utilize the NEO-Unify architecture, ensuring native 4K resolution generation and precise editing capabilities that allow for the reproduction of design frameworks in infographics and creative content.
The SenseNova U1.5-Lite-Preview model has only 8 billion MoT parameters, enabling it to generate high-density images while maintaining full detail even when scaled up, as well as creating artistic visuals and performing precise editing. In the process of content creation, complex tasks and edits determine whether the model can integrate into creative workflows.
The model is capable of processing high-density infographics that combine both content and layout. For instance, to depict a disassembled vintage camera, it needed to organize a timeline, knowledge modules, structure, and visual hierarchy within a single frame rendered in a detailed, monochrome engraved style.
The ability to reproduce design frameworks deserves special attention. When given an infographic about a journey through milestones of modern pop culture cinema and television, the model was able to transform it into a 'Mail Correspondence Journey' theme, preserving the curved path, node rhythm, and overall layout while replacing cinematic content with elements of delivery, sorting, transportation, and distribution. Thus, the model understood not only the style but also the entire design structure.
The editing functionality accounts for the reality of content production, which requires numerous revisions. In the case of a museum cultural product, the model maintained the specified structure, the MUSE inscription, the six-stage process, and the beige-brown color palette while simultaneously replacing the main visual object. This aligns with real changes where clients alter the theme and visual style but retain the overall structure.
In a complex test task, the model edited the hieroglyph 'ying' to 'ming', using the original ship hull as a final touch and restructuring the waves into remaining elements so that the new hieroglyph looked natural, as if growing from the sea. The model is also capable of considering multiple sources of information: in the fried chicken example, one image provided the recipe layout and style, while another provided the actual image of the food.
The release of SenseNova U1.5-Lite-Preview with open source strengthens the Chinese ecosystem of open-source multimodal models. Thanks to its lightweight profile with 8B-MoT parameters, local deployment becomes practical for developers, and native 4K support targets professional use cases, including advertising, editorial design, information design, and content production.
The unified NEO-Unify architecture eliminates the traditional separation between understanding and generation models, allowing a single model to interpret complex multimodal instructions and perform both creation and precise correction. The demonstrated capabilities of framework reproduction and context-preserving editing solve real problems in creative workflows, where the quality of edits and linguistic design consistency are as important as initial generation. This release positions SenseTime in the competitive landscape of image generation alongside closed and open models, offering developers a differentiated alternative through the combination of open weights, 4K output, lightweight deployment, and editing accuracy.
Lightelligence announced the achievement of a commercialization milestone for optical computing by introducing the PACE 3 chip featuring a 256x256 photonic matrix. At the WAIC 2026 forum, the company demonstrated the world's first deployed photonic computation in real-world conditions using the Tianshu Light Cube access control system.
The PACE 3 chip is specifically designed for performing inference on large models, not for training them. It utilizes a photonic in-memory computing architecture to execute matrix-vector multiplications, which is the primary computational pattern during neural network inference. This architecture provides lower latency and better energy efficiency compared to electronic counterparts.
The hybrid photo-electric architecture addresses a fundamental problem: while photonic computing offers advantages such as high throughput, reduced transmission losses, and inherent parallel processing capability, its practical application requires integration with electronic systems for data conversion and control. The main obstacle to scaling was a contradiction: end-users saw no incentive to reconfigure software for unproven hardware, and suppliers lacked market demand to cover the costs of producing optical devices.
Lightelligence's method involves combining the capabilities of optical switching, optical interconnects, and the photonic computations themselves within its product line. This allows for the creation of integrated photo-electric solutions that do not require clients to completely overhaul their software stack. Co-founder and CTO, Dr. Meng Huaiyu, noted that the size of the photonic matrix was intentionally chosen to achieve an optimal balance, as large computations require effective coordination between matrices of different sizes, rather than one excessively large one.
Dr. Shen Yichen compared the current stage of development to creating a hardware lottery for future applications, drawing parallels with how NVIDIA's gaming processors spurred the development of neural networks. Lightelligence also places simulators in classrooms so that next-generation algorithms can evolve on photonic hardware. The company asserts that photonic computing is not a replacement for GPUs but forms an entirely new ecosystem, and the demonstration of the access control system confirms long-term operational stability.
