SenseTime Launches Open-Source 9-Billion-Parameter AI Model for Enhanced Processing

SenseTime has released SenseNova Loop-9B-Preview, an open-source multimodal model designed to enhance processing capabilities by reusing computation layers without duplicating model weights. This announcement was made on October 30.

The model is capable of processing both text and images, generating text for various applications such as document understanding, coding, and tasks that involve external tools. It achieves deeper computation by running a shared block of 24 decoder layers twice, resulting in a model with approximately nine billion parameters.

Additionally, the training setup supports sequences of up to 262,144 tokens, integrating text, images, code, and examples of tool usage. SenseTime has made the model weights, inference instructions, and training guidance publicly available, distributing the project under the Apache 2.0 license.

For more details, the project can be accessed on GitHub at SenseTime’s repository.

Source: TechNode (China)