It has been exactly two months since Tencent initially launched the POINTS1.0 model. Recently, Tencent unveiled another major update—the official release of POINTS1.5. Building upon the classic LLaVA architecture of POINTS1.0, this next-generation model has undergone extensive performance optimizations and enhancements, providing users with an enhanced experience.

POINTS1.5 continues to adopt the foundational architecture from POINTS1.0, comprising a vision encoder, a projector, and a large language model. This enduring framework ensures that POINTS1.5 operates efficiently while fully harnessing the collaborative power of its components, thereby boosting overall performance.

According to Tencent’s official statement, POINTS1.5 significantly enhances model performance while maintaining the efficiency-first approach of POINTS1.0. This advancement allows POINTS1.5 to stand out in the global open-source model landscape, particularly among models with fewer than 10 billion parameters. Notably, the POINTS1.5-7B model has secured the top position thanks to its exceptional performance, surpassing many leading industry models such as Qwen2-VL, InternVL2, and MiniCPM-V-2.5.

In practical applications, POINTS1.5 also demonstrates robust capabilities. Whether in complex OCR tasks, reasoning, key information extraction, LaTeX formula extraction, mathematical processing, image translation, or object recognition, POINTS1.5 delivers impressive performance. This comprehensive and powerful application proficiency ensures that POINTS1.5 has broad potential across various industry scenarios.

Tencent has stated that the release of POINTS1.5 marks another significant milestone in the company’s ongoing efforts to advance artificial intelligence technology. Moving forward, Tencent will continue to increase its investment in the AI field, consistently introducing more advanced and practical models and technologies to provide users with higher quality and more convenient service experiences.