模型发布/更新
-
Anthropic releases Claude Opus 5
Anthropic has released Claude Opus 5, whose intelligence level approaches Claude Fable 5 at half the price. The model…
-
Ant Ling Releases Ling-3.0-flash Native Hybrid Reasoning Model
Ant Ling released Ling-3.0-flash, a new-generation native hybrid reasoning model with 124B total parameters and only…
-
Black Forest Labs Releases FLUX 3 Multimodal Foundation Model, Partners with mimic on FLUX-mimic Robot Model
Black Forest Labs released FLUX 3, a multimodal foundation model jointly trained on image, video and audio, with video…
-
Midjourney V8.2 released: focused on aesthetic improvements and personalized understanding
Midjourney today launched the V8.2 image model, focusing on improving aesthetic quality, image creativity, and…
-
Black Forest Labs 发布 FLUX 3 多模态模型,支持单次生成 20 秒视频与原生音频
Black Forest Labs 以 Early Access 方式推出 FLUX 3 多模态基础模型,采用统一架构联合学习图像、视频和音频。该模型基于 Self-Flow 学习框架扩展,可在单次生成中输出最长 20 秒视频并附带原生音…
产品发布/更新
-
Baidu Dazi Update: PC-Mobile Handoff, Desktop Built-in Browser, Cross-Device Continuous Execution of Complex Tasks
Baidu Dazi unveiled several upgrades at its recent AI Day, supporting PC-mobile interconnection with synchronized task…
-
OpenRouter Launches Classifiers Beta: Automatically Tagging AI Request Purpose and Cost Attribution
OpenRouter launched the Classifiers beta, allowing users to automatically tag each AI request's task type, department…
-
Claude Code v2.1.219 发布:新增 Claude Opus 5,支持 1M 上下文与嵌套子智能体
Claude Code v2.1.219 新增 Claude Opus 5 作为默认 Opus 模型,支持 1M 上下文窗口,快速模式定价为 $10/$50 每百万 token。
行业动态
-
Microsoft Outlines Path for Open Models to Boost US Competitiveness
Open-weight models are essential to a healthy AI ecosystem. Together with industry peers, we are charting a path for…
-
Kimi K3 Trails US Frontier Models Sharply in Cybersecurity Exploit Test, Knowledge Distillation May Be the Cause
A joint evaluation by the UK AI Security Institute and the US Center for AI Standards and Innovation shows that…
论文研究
-
Anthropic and Andon Labs release Drone-Bench to evaluate AI models' ability to autonomously pilot drones for locating and tracking targets
Anthropic has partnered with Andon Labs to launch Drone-Bench, which tests AI models' ability to autonomously pilot…
-
Apple 提出 LEAD 方法,破解长程推理中的“不可恢复瓶颈”
Apple 研究发现,大语言模型在长程执行中即使有高层策略也不稳定,极端分解会导致“不可恢复瓶颈”——少数“困难”步骤上的持续错误变得不可逆转。为此提出 Lookahead-Enhanced Atomic Decomposition(LE…
技巧与观点
-
New context engineering rules for Claude 5 generation models: Claude Code system prompt cut by over 80%
Anthropic removed more than 80% of Claude Code's system prompt for new-generation models such as Claude Opus 5 and…
-
Claude Design 产品设计师 Nate Parrott 分享如何用其探索视觉创意
Anthropic 产品设计师 Nate Parrott 分享了他在 Claude Design(beta 版)中利用 HTML 交互能力进行产品原型、幻灯片、动画等视觉创意探索与迭代的方法。他将 Anthropic 品牌规范提炼为提示词…
-
Claude 模型家族详解:如何为工作负载选择最佳模型
Anthropic 发布 Claude 模型选择指南,将模型分为 Mythos/Fable(最强能力,用于前沿编码与智能体任务)、Opus(推理密集型企业任务)、Sonnet(日常通用任务)和 Haiku(最低成本与最快速度)四个类别。