据了解，腾讯混元图像2.0模型支持文生图和绘画生图等多种交互方式。输入文字指令或上传本地图片、在线绘图，甚至只要一句话， 就 能毫秒级获得高质感图像。 【导读】 腾讯"AI全家桶"集中亮相世界人工智能大会，让"好用的AI"成为普惠生产力 中国基金报记者 张燕北 全球人工智能领域最具影响力的专业盛会 —— 世界人工智能大会（WAIC）7月26日在上海开幕。 连续第八 次 参加大会的腾讯，今年以"让'好用的AI'成为普惠生产力"为核心主题，携"AI全家桶"参展。其中多项技术应用"首发""首秀"， 集中展现腾讯如何凭借全自研云技术及混元大模型技术底座，让AI智能体成为14亿用户的"数字好友"，使人工智能技术沉淀为千行百业 的"新质生产力"。 混元多模态AI模型矩阵集中亮相，立体呈现"从云到端"自研链路 "元宝你好，帮我看看这几个玩偶分别是哪些品牌？""好的，这几个公仔分别是QQ企鹅公仔、Labubu和卡皮巴拉公仔……"在腾讯展台， 参观者拿起手机，和腾讯元宝开启一场视频聊天。 这是腾讯元宝最新推出的视频通话功能在公开场合的"首秀"。如今，腾讯元宝能够实现更低延迟、更拟人化、更有情感的视频和语音通 话。无论是老年人还 ...

Group 1 - The core viewpoint of the news highlights the performance of the Sci-Tech Innovation Board AI Index and the related ETF, indicating a positive trend in the AI sector with significant stock movements and liquidity [2][3] - The Sci-Tech AI ETF closely tracks the Sci-Tech Innovation Board AI Index, which consists of 30 large-cap companies providing foundational resources, technology, and application support for AI [3] - Recent developments in AI platforms, such as Manus and Tencent's new model, indicate ongoing innovation and competition in the AI field, with a focus on enhancing capabilities and user experience [2] Group 2 - The Sci-Tech AI ETF has seen a notable increase in scale, with a growth of 35.36 million yuan over the past week, ranking second among comparable funds [3] - The ETF's recent performance shows a net inflow of 11 million yuan over the last five trading days, indicating strong investor interest [4] - The ETF's historical performance includes a maximum monthly return of 15.59% and a 60% probability of profit in up months, showcasing its potential for returns [4] Group 3 - The top ten weighted stocks in the Sci-Tech Innovation Board AI Index account for 70.68% of the index, with notable companies like Lanqi Technology and Cambricon leading the list [6][8] - The ETF has a management fee of 0.50% and a custody fee of 0.10%, which are among the lowest in comparable funds, indicating cost efficiency for investors [6] - The tracking error of the ETF over the past two months is 0.013%, reflecting high tracking precision compared to similar funds [6]

人像摄影风格 动物特写 AI图像生成进入"毫秒级"时代。 5月16日，腾讯发布最新混元图像2.0模型（Hunyuan Image2.0），基于模型架构创新，在行业内率先实现实时生图，在画面质感超写实的基础上， 带来全新的AI生图交互体验。模型于即日起在腾讯混元官方网站上线，并对外开放注册体验。 相比前代模型，腾讯混元图像2.0模型参数量提升了一个数量级，得益于超高压缩倍率的图像编解码器以及全新扩散架构，其生图速度显著快于行业领 先模型，在同类商业产品每张图推理速度需要5到10秒的情况下，腾讯混元可实现毫秒级响应，支持用户可以一边打字或者一边说话一边出图，改变了 传统"抽卡—等待—抽卡"的方式，带来交互体验革新。 除了速度快以外，腾讯混元图像2.0模型图像生成质量提升明显，通过强化学习等算法以及引入大量人类美学知识对齐，生成的图像可有效避免AIGC图 像中的"AI味" ，真实感强、细节丰富、可用性高。 在图像生成领域专门测试模型复杂文本指令理解与生成能力的评估基准 GenEval（Geneval Bench）上，腾讯混元图像2.0模型准确率超过95%，远超其他同类模型。 | | | | | | | 文生图客观指 ...