AI在线 AI在线

Model

Google Launches AlphaGenome: AI Technology Helps Cracking the Code of Life in One Click

Google Launches AlphaGenome: AI Technology Helps Cracking the Code of Life in One Click
In the field of life sciences, Google DeepMind has once again set a new trend by introducing a revolutionary artificial intelligence tool - AlphaGenome. This new AI model can not only read up to 1 million DNA bases but also accurately predict how gene mutations affect molecular functions, marking a major breakthrough in biological research.  Decoding the Blueprint of LifeThe genome is the blueprint of life, recording the DNA instructions for each cell.
6/26/2025 5:01:39 PM
AI在线
Silicon Base Flow Launches the World's First Open Source Large-scale Hybrid Attention Reasoning Model MiniMax-M1-80k

Silicon Base Flow Launches the World's First Open Source Large-scale Hybrid Attention Reasoning Model MiniMax-M1-80k
The Silicon Cloud (SiliconFlow) has officially launched the world's first open-source large-scale hybrid attention reasoning model — MiniMax-M1-80k (456B). This innovative model is designed to provide strong support for complex tasks such as software engineering, long-context understanding, and tool usage, and its performance can rival leading models like o3 and Claude4Opus.It is reported that MiniMax-M1-80k supports a maximum context length of up to 128K, greatly facilitating the handling of long texts. For users with special needs, the platform also provides backend support to meet the demand for 1M long contexts.
6/17/2025 9:03:21 PM
AI在线
Former Google CEO's startup releases 24 billion-parameter chemical reasoning model with accuracy surpassing multiple leading models

Former Google CEO's startup releases 24 billion-parameter chemical reasoning model with accuracy surpassing multiple leading models
In the field of artificial intelligence, research on large models continues to advance, particularly in improving reasoning capabilities. Recently, FutureHouse, a startup funded by former Google CEO Eric Schmidt, has open-sourced a chemical task reasoning model named ether0, with a parameter scale as high as 24 billion. This model demonstrates strong domain-specific capabilities in chemistry without requiring additional pre-training in specific fields, achieving remarkable results through post-training techniques while significantly reducing data requirements compared to traditional field-specific models.The application of reasoning models goes beyond simple multiple-choice tests.
6/17/2025 9:03:21 PM
AI在线
Kimi-Dev-72B: The AI Wonder Breaking the Boundaries of Code Repair

Kimi-Dev-72B: The AI Wonder Breaking the Boundaries of Code Repair
Recently, the much-anticipated open-source large language model Kimi-Dev-72B has officially launched, becoming a favorite among developers. This model was developed by the "Dark Side of the Moon" team and is specifically designed to solve coding problems, aiming to enhance programming efficiency.In the recent SWE-bench Verified test, Kimi-Dev-72B demonstrated extraordinary capabilities, particularly excelling in fixing code defects within Docker environments. This advantage makes Kimi-Dev-72B not only a valuable assistant for developers but also an important tool for optimizing development workflows.The core advantage of this model lies in its reinforcement learning-based optimization mechanism.
6/17/2025 9:03:21 PM
AI在线
DeepSeek R1 Model Shocks the AI World: Low-Cost, High Efficiency Leads a New Industry Track

DeepSeek R1 Model Shocks the AI World: Low-Cost, High Efficiency Leads a New Industry Track
In January of this year, the release of DeepSeek's R1 model was not just an ordinary AI announcement; it was hailed as a "watershed moment" in the tech industry, causing a significant stir across the entire technology sector and forcing industry leaders to rethink their fundamental approaches to AI development. DeepSeek's extraordinary achievements did not stem from novel features but from its ability to deliver results comparable to those of tech giants at a fraction of the cost, marking the rapid progress of AI along two parallel tracks: "efficiency" and "computing."Innovation Under Constraints: High Performance at Low CostDeepSeek's emergence has been remarkable, showcasing the capability for innovation even under significant constraints. In response to U.S.
6/16/2025 12:01:13 PM
AI在线
Video Version of AI Clothes Swapping Framework MagicTryOn Based on Wan2.1 Video Model

Video Version of AI Clothes Swapping Framework MagicTryOn Based on Wan2.1 Video Model
In the modern fashion industry, Video Virtual Try-On (VVT) has gradually become an important component of user experience. This technology aims to simulate the natural interaction between clothing and human body movements in videos, showcasing realistic effects during dynamic changes. However, current VVT methods still face multiple challenges such as spatial-temporal consistency and preservation of clothing content.To address these issues, researchers proposed MagicTryOn, a virtual try-on framework based on a large-scale video diffusion transformer (Diffusion Transformer).
6/16/2025 12:01:13 PM
AI在线
Ant Group and inclusionAI Jointly Launch Ming-Omni: The First Open Source Multi-modal GPT-4o

Ant Group and inclusionAI Jointly Launch Ming-Omni: The First Open Source Multi-modal GPT-4o
Recently, Inclusion AI and Ant Group jointly launched an advanced multimodal model called "Ming-Omni," marking a new breakthrough in intelligent technology. Ming-Omni is capable of processing images, text, audio, and video, providing powerful support for various applications. Its functions not only cover speech and image generation but also possess the ability to integrate and process multimodal inputs.** Comprehensive Multimodal Processing Capability **.
6/16/2025 11:01:43 AM
AI在线
Xiaohongshu makes a major move! The all-new open-source large model dots.llm1震撼登场 with 142 billion parameters!

Xiaohongshu makes a major move! The all-new open-source large model dots.llm1震撼登场 with 142 billion parameters!
Recently, the hi lab team of Xiaohongshu officially released its first open-source text large model — dots.llm1. This new model has attracted extensive attention in the industry due to its outstanding performance and massive number of parameters.dots.llm1 is a large-scale Mixture of Experts (MoE) language model with an impressive 142 billion parameters, including 14 billion activated parameters. After being trained on 11.2 TB of high-quality data, this model's performance can rival Alibaba's Qwen2.5-72B.
6/16/2025 9:48:52 AM
AI在线
Former DeepSeek executive secretly starts new AI Agent project, already backed by top VC

Former DeepSeek executive secretly starts new AI Agent project, already backed by top VC
According to the news from Tiger嗅, a core executive of the domestic large model company DeepSeek has quietly left and started a new business half a year ago, and plans to launch its first Agent product around Christmas in 2025.Sources close to the matter told Tiger嗅 that the executive once held the role of "CTO" at DeepSeek. However, some insiders pointed out that DeepSeek's internal structure does not clearly set up a "CTO" position. In name, this position may not exist, but there is indeed an executive who takes on technical coordination and R&D decision-making responsibilities similar to those of a CTO.Reliable sources also revealed that this startup project has already received financing support from a leading VC, with the specific amount undisclosed.
6/16/2025 9:48:51 AM
AI在线
The world's first female tumor AI large model, Mulan, is online with free mobile service!

The world's first female tumor AI large model, Mulan, is online with free mobile service!
Recently, Huazhong University of Science and Technology announced that the world's first female tumor artificial intelligence large model, "Mulan," has officially entered the clinical application stage. This significant medical technology was jointly developed by the National Clinical Research Center for Obstetric Diseases at Tongji Hospital, affiliated with the Tongji Medical College of Huazhong University of Science and Technology, and multiple institutions. Its aim is to enhance the screening and treatment levels of female tumors..
5/15/2025 10:01:58 AM
AI在线
OpenAI releases the new GPT-4.1 model, more proficient in programming tasks

OpenAI releases the new GPT-4.1 model, more proficient in programming tasks
In the ongoing innovation in the field of artificial intelligence, OpenAI recently announced a significant upgrade to its ChatGPT chatbot by launching the latest GPT-4.1 model. Starting from May 14th, this model has been officially available to users, offering new options for Pro, Plus, and Team users. Meanwhile, Enterprise and Edu users will also gain access within the coming weeks, ensuring that more users can experience this advanced technology.The launch of the GPT-4.1 model marks another leap forward for OpenAI in handling programming tasks.
5/15/2025 10:01:58 AM
AI在线
Google DeepMind Launches AlphaEvolve: AI Breaks a 56-Year Record in Mathematics and Optimizes Its Own Training System

Google DeepMind Launches AlphaEvolve: AI Breaks a 56-Year Record in Mathematics and Optimizes Its Own Training System
Google DeepMind today released AlphaEvolve, an artificial intelligence agent with self-evolution capabilities that can independently invent complex computer algorithms and has been widely applied in Google's data centers, chip design, and AI model training, achieving significant results.AlphaEvolve combines the Gemini large language model with evolutionary optimization methods to automatically test, improve, and enhance the entire codebase, not just a single function. This system has quietly run internally for over a year, improving computing resource scheduling efficiency, accelerating model training, and achieving breakthroughs in mathematical research.From Servers to Chips: AlphaEvolve Optimizes Google's Underlying ArchitectureThe scheduling algorithm proposed by AlphaEvolve has already been deployed in Google's global data centers, addressing the "resource stranded" problem and recovering 0.7% of resources.
5/15/2025 10:01:58 AM
AI在线
The Next Generation Open Source 3D Model Step1X-3D Debuts, AI Industry Trend Draws Attention

The Next Generation Open Source 3D Model Step1X-3D Debuts, AI Industry Trend Draws Attention
Recently, the technology sector welcomed a brand-new open-source 3D large model called "Step1X-3D." The release of this model marks another significant advancement in AI technology, particularly in 3D modeling and reasoning capabilities. Not only is this model open-source, but it also provides developers with various practical features, greatly promoting innovation and research possibilities.At the same time, Xiaomi is continuously expanding its presence in the AI field. It has recently applied for the "MiMo" trademark, which is intended to be used for inference large models.
5/15/2025 10:01:53 AM
AI在线
Alibaba Open Sources All-in-one Video Foundation Model to Empower Video Generation and Editing

Alibaba Open Sources All-in-one Video Foundation Model to Empower Video Generation and Editing
On the evening of May 14th, Alibaba officially launched Tongyi Wanxiang Wan2.1-VACE, which is currently the most comprehensive video generation and editing model in the industry. The highlight of this model lies in its multiple powerful capabilities, enabling it to simultaneously achieve text-to-video generation, image-based video generation, video retouching, local editing, background extension, duration extension, and other foundational generation and editing functions. This innovative product further lowers the threshold for video production, allowing more creators to easily get started..
5/15/2025 10:01:52 AM
AI在线
Adobe 推出全新 AI 视频生成器 Firefly Video Model，完全使用授权内容进行训练

Adobe 推出全新 AI 视频生成器 Firefly Video Model，完全使用授权内容进行训练
Adobe 公司今日发布了全新的人工智能驱动的文本转视频工具 Firefly Video Model。该工具能够根据文本提示生成全新的视频，与竞争对手不同，Adobe 声称 Firefly Video Model 完全使用授权内容进行训练，有望规避其他生成式 AI 工具所面临的伦理和版权问题。AI在线注意到，由于其使用授权内容进行训练，Adobe 称 Firefly Video Model 是“第一个公开可用的商业安全视频模型”。
10/15/2024 7:14:23 AM
远洋
Meta SAM 2 登场：首个能在图片和视频中实时分割对象的统一开源 AI 模型

Meta SAM 2 登场：首个能在图片和视频中实时分割对象的统一开源 AI 模型
感谢Meta 公司发布 Meta Segment Anything Model 2（SAM2），SAM 2 能分割任何目标，能在一个视频中实时追踪所有镜头 —— 解锁新的视频编辑能力并在混合现实中提供新的体验。Meta 公司今天发布新闻稿，介绍了全新的 Meta Segment Anything Model 2（SAM 2）模型，先支持分割视频和图像中的对象。开源Meta 公司宣布将以 Apache 2.0 许可发布 SAM 2，因此任何人都可以使用它来构建自己的体验。Meta 还将以 CC BY 4.0 许可共享
7/30/2024 9:58:28 AM
故渊
ICML 2024 | 特征污染：神经网络会学习不相关特征而泛化失败

ICML 2024 | 特征污染：神经网络会学习不相关特征而泛化失败
论文标题：Feature Contamination: Neural Networks Learn Uncorrelated Features and Fail to Generalize论文链接：：，深度神经网络 SGD scaling的机器学习范式再次证明了其在AI领域的主导地位。为什么基于深度神经网络的范式能够取得成功？比较普遍的观点是：神经网络具有从海量的高维输入数据中自动学习抽象而可泛化的特征的能力。遗憾的是，受限于当前分析手段和数学工具的不足，目前我们对于“（深度）神经网络如何实现这样的特征学习过程”这
6/24/2024 10:44:00 AM
新闻助手
OpenAI 推出 Model Spec 拟议框架，探索生成 NSFW 内容等 AI 响应规范

OpenAI 推出 Model Spec 拟议框架，探索生成 NSFW 内容等 AI 响应规范
OpenAI 公司近日发布名为 Model Spec 的拟议框架初稿，希望能够规范 AI 模型和工具（例如 GPT-4）未来的响应方式。OpenAI 提出了三个普适原则：AI 模型应遵循相关规范，协助开发者和终端用户并提供有用响应考虑潜在利益和危害的情况下造福人类响应应符合社会道德或者相关法律OpenAI 还提出了其它几条原则：遵循指挥系统遵守适用法律不提供信息危害尊重创作者及其权利保护人们的隐私不回复 NSFW（工作场所不宜）内容OpenAI 表示之所以倡导并推行该框架，其目的是让企业和用户切换 AI 模型的“辛
5/9/2024 8:58:47 AM
故渊
  • 1