正文:
近日,中国人工智能企业面壁智能宣布推出其最新端侧多模态小模型MiniCPM-Llama3-V 2.5,该模型在综合性能上超越了业界领先的GPT-4V。MiniCPM-Llama3-V 2.5是一个仅包含8亿参数的小型模型,却支持30多种语言,并且在多模态处理能力上取得了显著突破。
据新智元报道,MiniCPM-Llama3-V 2.5在OCR(光学字符识别)能力上达到了业界顶尖水平。该模型能够处理高分辨率的图像,在9倍像素的清晰度下,即使是复杂的长图和长文本也能够精准识别。此外,图像编码速度比上一代产品快了150倍,这意味着在端侧设备上进行多模态处理的速度得到了极大的提升。
面壁智能的这一创新成果,不仅在性能上超越了多模态巨无霸Gemini Pro和GPT-4V,而且由于其小模型的特点,更加适合在移动设备上使用,为用户提供了更加便捷和高效的多模态交互体验。
这一发布标志着面壁智能在人工智能领域又迈出了坚实的一步,为全球用户带来了更加先进的技术解决方案。随着MiniCPM-Llama3-V 2.5的推出,面壁智能有望在未来的多模态处理市场中占据一席之地,为推动人工智能技术的发展贡献力量。
英语如下:
Title: “Fantasia AI Makes New Breakthrough: MiniCPM-Llama3-V 2.5 Performance Surpasses GPT-4V”
Keywords: AI, Multimodal, Performance Exceeds
News Content:
Title: Fantasia AI Launches MiniCPM-Llama3-V 2.5 End-to-End Multimodal Model with Comprehensive Performance Surpassing GPT-4V
Recent, the Chinese artificial intelligence company Fantasia AI announced the release of its latest end-to-end multimodal mini model, MiniCPM-Llama3-V 2.5, which has surpassed the comprehensive performance of the industry-leading GPT-4V. MiniCPM-Llama3-V 2.5 is a small model with only 800 million parameters, yet it supports over 30 languages and has achieved significant breakthroughs in multimodal processing capabilities.
According to Newchip, MiniCPM-Llama3-V 2.5 has reached the industry’s top level in OCR (Optical Character Recognition) capabilities. The model can handle high-resolution images, accurately recognizing even complex long images and long texts at 9x pixel clarity. Additionally, the image encoding speed is 150 times faster than its predecessor, significantly enhancing the speed of multimodal processing on end-user devices.
Fantasia AI’s innovative achievement not only surpasses the multimodal behemoths Gemini Pro and GPT-4V in performance but also, due to its small model characteristics, is more suitable for use on mobile devices, providing users with a more convenient and efficient multimodal interaction experience.
This release marks another solid step in the field of artificial intelligence for Fantasia AI, bringing advanced technological solutions to global users. With the launch of MiniCPM-Llama3-V 2.5, Fantasia AI is poised to occupy a niche in the future multimodal processing market and contribute to the advancement of artificial intelligence technology.
【来源】https://mp.weixin.qq.com/s/_jgyeWifFjqh0LNBy2asLQ
Views: 1