AI 信息雷达 | Anaxagore

AI 信息雷达 2026-11-06

Fri, 06 Nov 2026 00:00:00 +0000

必看

2026 年 PyTorch 大会：谷歌和 Meta 开源 TorchTPU，在 TPU 上原生支持 PyTorch — 点击查看原文>
当 Coding Agent 走进研发团队：我们打造并开源了自进化研发平台 LoopForge — 点击查看原文>
突发：李飞飞最新AI成果发布，把世界模型装进复古掌机，单张图实时跑出15帧 - 凤凰网科技 — 突发：李飞飞最新AI成果发布，把世界模型装进复古掌机，单张图实时跑出15帧凤凰网科技
长视频生成模型为什么越跑越"呆"：LongTake如何让AI视频保持生命力 - 科技行者 — 长视频生成模型为什么越跑越"呆"：LongTake如何让AI视频保持生命力科技行者
谷歌发布Gemini Omni 1.2 Flash视频生成模型 - 维度网 — 谷歌发布Gemini Omni 1.2 Flash视频生成模型维度网
World Labs推出Frameboy：不用预设镜头，也能实时探索AI生成的三维世界 - 凤凰网 — World Labs推出Frameboy：不用预设镜头，也能实时探索AI生成的三维世界凤凰网
Anthropic’s new AI model achieves over 50% on all benchmarks in METR analysis - Crypto Briefing — Anthropic’s new AI model achieves over 50% on all benchmarks in METR analysis Crypto Briefing
Hungary: Communist-era files released, up to 60k former agents revealed —
Zachary Perret: Anthropic and Claude partner to launch U.S. personal finance features - Traders Union — Zachary Perret: Anthropic and Claude partner to launch U.S. personal finance features Traders Union
Google Launches Gemini Omni 1.2 Flash with 15-Second Video Generation - Blockchain News — Google Launches Gemini Omni 1.2 Flash with 15-Second Video Generation Blockchain News

值得看

Anthropic Warns AI Could Soon Accelerate Its Own Development - Cirebon Times — Anthropic Warns AI Could Soon Accelerate Its Own Development Cirebon Times
NVIDIA Cosmos 3 Simulation Models Add Interactive Control to Generated Worlds - remio — NVIDIA Cosmos 3 Simulation Models Add Interactive Control to Generated Worlds remio
Dreamina AI Announces Guide to AI Video Generation for Creative Projects - The Manila Times — Dreamina AI Announces Guide to AI Video Generation for Creative Projects The Manila Times
云端一天抵人间三年，拆解萝卜快跑的世界模型秘密 - 汽车之家 — 云端一天抵人间三年，拆解萝卜快跑的世界模型秘密汽车之家
都在做世界模型，自动驾驶的真实数据会不会越来越不值钱？ - 电子工程专辑EE Times China — 都在做世界模型，自动驾驶的真实数据会不会越来越不值钱？电子工程专辑EE Times China
Open Models Crowd the Top as Kimi and Mistral Match the Frontier - PYMNTS.com — Open Models Crowd the Top as Kimi and Mistral Match the Frontier PYMNTS.com
InVideo AI 评测：最适合创作者的 AI 视频生成器 - quasa.io — InVideo AI 评测：最适合创作者的 AI 视频生成器 quasa.io
【视频】大模型芯片+世界模型上车2027款风云T11将成软硬一体智驾首搭车型 - 汽车之家 — 【视频】大模型芯片+世界模型上车2027款风云T11将成软硬一体智驾首搭车型汽车之家
Google Officially Announces “Gemini Omni 1.2 Flash” Video Generation AI - jetstream.blog — Google Officially Announces “Gemini Omni 1.2 Flash” Video Generation AI jetstream.blog
Cult.fit builds operational muscle with Codex — Cult.fit is using Codex to help teams across functions build tools, move faster, and change how work gets done.

AI 信息雷达 2026-11-05

Thu, 05 Nov 2026 00:00:00 +0000

必看

Kimi Hosted Agents 正式发布：Harness as a Service - 53AI — Kimi Hosted Agents 正式发布：Harness as a Service 53AI
魔芯科技11月4日发布昇腾原生4D世界模型MoWorld2.0 - 观点网 — 魔芯科技11月4日发布昇腾原生4D世界模型MoWorld2.0 观点网
2.8 万亿参数 Kimi K3.1 Pro 开放权重，登顶 Agent Arena 开源模型榜 - DoNews — 2.8 万亿参数 Kimi K3.1 Pro 开放权重，登顶 Agent Arena 开源模型榜 DoNews
月之暗面开源Kimi K3.1 Pro模型权重：登顶Agent Arena开放模型榜首，Bash可靠性跃居第一 - SmartHey — 月之暗面开源Kimi K3.1 Pro模型权重：登顶Agent Arena开放模型榜首，Bash可靠性跃居第一 SmartHey
月之暗面开源Kimi K3.1 Pro，登顶Agent Arena开放权重模型榜 - 凤凰网 — 月之暗面开源Kimi K3.1 Pro，登顶Agent Arena开放权重模型榜凤凰网
谷歌 Data Agent Kit 正式发布：安全防护分为 Skill 与 IAM — 点击查看原文>
PyTorch 2.15’s live Q&A is tomorrow, November 5, at 11 AM PT. Bring your questions about the release, which makes the re — PyTorch 2.15’s live Q&A is tomorrow, November 5, at 11 AM PT. Bring your questions about the release, which makes the rewritten NCCL backend the default and…
Ripple-65B Preview: U.S. Open-Weight SOTA on Agentic Coding —
启元具身研究中心发布世界模型仿真器PrimeSim，权重、代码将开源_财经报道_财经 - 证券之星 — 启元具身研究中心发布世界模型仿真器PrimeSim，权重、代码将开源_财经报道_财经证券之星
英伟达团队下场！国产视频模型开源，硅谷公司上门谈授权 - 凤凰网科技 — 英伟达团队下场！国产视频模型开源，硅谷公司上门谈授权凤凰网科技

值得看

让2000名工程师使用中国大模型，这家英国企业实现了“词元自由” - 凤凰网 — 让2000名工程师使用中国大模型，这家英国企业实现了“词元自由” 凤凰网
基于本体+知识+Harness+Datat Agent的智能体完整落地链路！ — Data + AI 技术实践巡展 · 上海站
Baidu Unveils Apollo World, a World Model Designed to Let AI Train AI - Gasgoo — Baidu Unveils Apollo World, a World Model Designed to Let AI Train AI Gasgoo
Baidu unveils Apollo World model to let AI teach AI to drive - CnEVPost — Baidu unveils Apollo World model to let AI teach AI to drive CnEVPost
百度无人车搞出“黑客帝国”，世界模型实时推演全局，跑一天抵三年 - 区块周刊BlockWeeks — 百度无人车搞出“黑客帝国”，世界模型实时推演全局，跑一天抵三年区块周刊BlockWeeks
从3D生成到世界模型，这家AI公司想和年轻人一起定义问题 — 薪资+期权不限！实习生也能训核心模型
【视频】世界模型为什么是无人驾驶规模化的核心底牌？ - 汽车之家 — 【视频】世界模型为什么是无人驾驶规模化的核心底牌？汽车之家
What is driving the unprecedented acceleration in AI? Mark Collier summed it up in one word: OPEN. At the center of this — What is driving the unprecedented acceleration in AI? Mark Collier summed it up in one word: OPEN…
可灵AI宣布Kling 4.0 Flash全面开放覆盖官网、API及Adobe系列插件 - 观点网 — 可灵AI宣布Kling 4.0 Flash全面开放覆盖官网、API及Adobe系列插件观点网
大晓机器人携手海光，完成世界模型国产算力适配 - www.gasgoo.com — 大晓机器人携手海光，完成世界模型国产算力适配 www.gasgoo.com

AI 信息雷达 2026-11-04

Wed, 04 Nov 2026 00:00:00 +0000

必看

中国开源AI大模型持续“圈粉” - 手机新浪网 — 中国开源AI大模型持续“圈粉” 手机新浪网
腾讯混元Hy4即将发布：预览版位列Agent Arena全球第十，任务成本比Kimi低68% - 凤凰网 — 腾讯混元Hy4即将发布：预览版位列Agent Arena全球第十，任务成本比Kimi低68% 凤凰网
百度发布Apollo World世界模型，引爆物理 AI，数据服务商受益丨大咖星选 - 搜狐网 — 百度发布Apollo World世界模型，引爆物理 AI，数据服务商受益丨大咖星选搜狐网
India’s CERT-In gets access to Anthropic’s unreleased Mythos AI model for cyber defence - CNBC TV18 — India’s CERT-In gets access to Anthropic’s unreleased Mythos AI model for cyber defence CNBC TV18
视频生成模型Seedance过去半年海外市场收入增近5倍｜即时新闻｜产经｜on.cc东网 - on.cc東網 — 视频生成模型Seedance过去半年海外市场收入增近5倍｜即时新闻｜产经｜on.cc东网 on.cc東網
国内三大顶尖大模型负责人乌镇同台对话：开源后如何防范风险 - ThePaper.cn — 国内三大顶尖大模型负责人乌镇同台对话：开源后如何防范风险 ThePaper.cn
中国蓝观察丨开源大模型为何成了浙江金名片？ - https://www.hzsc.com.cn/ — 中国蓝观察丨开源大模型为何成了浙江金名片？ https://www.hzsc.com.cn/
字节视频模型Seedance海外收入占比达50%字节跳动旗下视频生成模型Seedance的海外收入占… - 禁闻网 — 字节视频模型Seedance海外收入占比达50%字节跳动旗下视频生成模型Seedance的海外收入占… 禁闻网
百度Apollo在深圳发布ApolloWorld世界模型 - 观点网 — 百度Apollo在深圳发布ApolloWorld世界模型观点网
DeepSeek发布创作者模式，下一代大模型计划基于华为昇腾芯片预训练 - ThePaper.cn — DeepSeek发布创作者模式，下一代大模型计划基于华为昇腾芯片预训练 ThePaper.cn

值得看

百度披露Apollo World世界模型无人车虚拟世界路测1天抵现实3年 - 香港01 — 百度披露Apollo World世界模型无人车虚拟世界路测1天抵现实3年香港01
【视频】画得像只是世界模型最低门槛百度王亮：不能做生成炫酷视频的画师 - 汽车之家 — 【视频】画得像只是世界模型最低门槛百度王亮：不能做生成炫酷视频的画师汽车之家
CERT-In gets Mythos access; to assess its use in measured manner - ET Telecom — CERT-In gets Mythos access; to assess its use in measured manner ET Telecom
Seedance海外收入大增5倍版权合作推动增长 - 中华网 — Seedance海外收入大增5倍版权合作推动增长中华网
曝腾讯位列Seedance国内客户前三头部企业采购趋势显现 - 中华网 — 曝腾讯位列Seedance国内客户前三头部企业采购趋势显现中华网
After spending tens of billions on R&D, Anthropic sees reasons to slow its pursuit of self-improving AI - R&D World — After spending tens of billions on R&D, Anthropic sees reasons to slow its pursuit of self-improving AI R&D World
世界模型的"瘦身术"：如何让AI用更少的信息做出更聪明的规划 - 科技行者 — 世界模型的"瘦身术"：如何让AI用更少的信息做出更聪明的规划科技行者
【视频】AI教AI开车，百度世界模型打通无人驾驶不可能三角 - 汽车之家 — 【视频】AI教AI开车，百度世界模型打通无人驾驶不可能三角汽车之家
Anthropic IPO Filing Shows 85.4% Margin vs Z.ai and MiniMax - News and Statistics - IndexBox — Anthropic IPO Filing Shows 85.4% Margin vs Z.ai and MiniMax - News and Statistics IndexBox
字节跳动Seedance海外收入占比已达50%；苹果CEO回应iPhone Duo屏幕问题｜数智早参 - https://www.mrjjxw.com/ — 字节跳动Seedance海外收入占比已达50%；苹果CEO回应iPhone Duo屏幕问题｜数智早参 https://www.mrjjxw.com/

AI 信息雷达 2026-11-03

Tue, 03 Nov 2026 00:00:00 +0000

必看

DeepSeek发布Super-agent架构推动智能体向自主管家型演进 - 品玩 — DeepSeek发布Super-agent架构推动智能体向自主管家型演进品玩
【国际动态】刚刚，谷歌Genie 4发布！世界模型还得是谷歌 - 手机新浪网 — 【国际动态】刚刚，谷歌Genie 4发布！世界模型还得是谷歌手机新浪网
PyTorch Foundation project Ray is today announcing the SIG for Ray on Accelerators: a vendor-neutral roadmap for running — PyTorch Foundation project Ray is today announcing the SIG for Ray on Accelerators: a vendor-neutral roadmap for running Ray on NVIDIA, AMD, TPU, and custom…
Announcing the Special Interest Group (SIG) for Ray on Accelerators — Today, we are announcing the Special Interest Group (SIG) for Ray on hardware accelerators. When Ray was created in UC Berkeley over 10 years ago, the…
42AI Launches a Behavioral World Model to Predict Human Decisions - Unite.AI — 42AI Launches a Behavioral World Model to Predict Human Decisions Unite.AI
Introducing the Grid Protectors Program — OpenAI’s Grid Protectors Program will work with utilities to help future compute campuses support grid reliability and protect consumer electricity costs.
Import AI 479: Spacewar; low-latency tasks versus LLMs; Genie 4 — Welcome to Import AI, a newsletter about AI research. Import AI runs on arXiv, cappuccinos, and feedback from readers…
Athenahealth launches tool to bring patient health records into Anthropic's Claude - Fierce Healthcare — Athenahealth launches tool to bring patient health records into Anthropic's Claude Fierce Healthcare
Kimi发布游戏开发与创意工作解决方案，Agent操作专业引擎 - 凤凰网 — Kimi发布游戏开发与创意工作解决方案，Agent操作专业引擎凤凰网
DeepSeek发布智能体新架构：让AI管AI - 财联社 — DeepSeek发布智能体新架构：让AI管AI 财联社

值得看

框架选不对，Claude 开发成本多一倍：前端十年框架膨胀，终于有人买单了 — 点击查看原文>
模拟计算进入矩阵时代！北大联手硅谷IBM，写下新定义 — GPU迭代一百次的事，它一步做完
Anthropic Addresses Video Game Cloning With Minor Claude Update - CNET — Anthropic Addresses Video Game Cloning With Minor Claude Update CNET
Laura Shin: Anthropic research model found Linux kernel bug behind urgent Solana validator alert - Traders Union — Laura Shin: Anthropic research model found Linux kernel bug behind urgent Solana validator alert Traders Union
llm 0.37 — Release: llm 0.37 There's a lot of stuff in this new release of my LLM CLI tool and Python library, representing around two months of accumulated features and…
Building Legal Agents When You Can't See the Data - Jacob Lauritzen, CTO of Legora — Legora can’t read a single trace of what its agent does for a lawyer. The data belongs to the law firm’s clients…
Keep Distributed Training Running with Automated Workload Recovery — As distributed AI training scales, failures across hardware and software dependencies increasingly disrupt training progress, making fast, reliable recovery...
11月3日电商早报：阿里千问接入苹果智能；快手可灵AI增资约198亿 - 亿邦动力网 — 11月3日电商早报：阿里千问接入苹果智能；快手可灵AI增资约198亿亿邦动力网
快手可灵AI完成约198亿元增资交割- 快讯 - 亿邦动力网 — 快手可灵AI完成约198亿元增资交割- 快讯亿邦动力网
Speedrun AI Labs, Builder of Cybersecurity and Compliance Agents, Accepted Into Anthropic's Claude Startups Program - USA Today — Speedrun AI Labs, Builder of Cybersecurity and Compliance Agents, Accepted Into Anthropic's Claude Startups Program USA Today

AI 信息雷达 2026-11-02

Mon, 02 Nov 2026 00:00:00 +0000

必看

小K播早报|苹果和阿里巴巴达成合作 DeepSeek发布全新Agent架构 - Moomoo — 小K播早报|苹果和阿里巴巴达成合作 DeepSeek发布全新Agent架构 Moomoo
刚刚，DeepSeek重磅开源：Agent底座架构！ - 智源社区 — 刚刚，DeepSeek重磅开源：Agent底座架构！智源社区
Weeks Ahead Of Expected Anthropic IPO, DeepSeek Launches Autonomous Coding Agent To Take On Claude Code - Stocktwits — Weeks Ahead Of Expected Anthropic IPO, DeepSeek Launches Autonomous Coding Agent To Take On Claude Code Stocktwits
DeepSeek开源Super-agent研究原型：让AI主动发现Bug并持续参与日常工作 - 虎嗅网 — DeepSeek开源Super-agent研究原型：让AI主动发现Bug并持续参与日常工作虎嗅网
机器人行业研究：谷歌升级Genie 4世界模型，智元大会发布新一代G2 Max - 证券之星 — 机器人行业研究：谷歌升级Genie 4世界模型，智元大会发布新一代G2 Max 证券之星
DeepSeek 发布 Super-agent 开源原型 - Pasquale Pillitteri — DeepSeek 发布 Super-agent 开源原型 Pasquale Pillitteri
DeepSeek发布全新Agent架构 - 观点网 — DeepSeek发布全新Agent架构观点网
【视频】XHEART发布大模型芯片X7 Ultra，Momenta预告R7 Ultra世界模型 - 车家号 — 【视频】XHEART发布大模型芯片X7 Ultra，Momenta预告R7 Ultra世界模型车家号
DeepSeek发布Super-agent设计探索 AI指挥AI 智能体 - 香港01 — DeepSeek发布Super-agent设计探索 AI指挥AI 智能体香港01
DeepSeek发布全新Agent架构迈向更自主的AI系统 - 中华网 — DeepSeek发布全新Agent架构迈向更自主的AI系统中华网

值得看

Kuaishou's Kling AI completes $2.9B financing - TNGlobal — Kuaishou's Kling AI completes $2.9B financing TNGlobal
DeepSeek重构Agent！AI应用加速落地多股业绩倍增可期(名单)_股市聚焦_市场_中金在线 - 中金在线 — DeepSeek重构Agent！AI应用加速落地多股业绩倍增可期(名单)_股市聚焦_市场_中金在线中金在线
DeepSeek大动作！Agent商业化有望提速这14股业绩逐季高增(名单) - 东方财富 — DeepSeek大动作！Agent商业化有望提速这14股业绩逐季高增(名单) 东方财富
机器人学会的世界模型，为什么到了新环境就"失灵"了？ - 科技行者 — 机器人学会的世界模型，为什么到了新环境就"失灵"了？科技行者
Anthropic says you can't be mean to Claude. It's not about feelings - KSL News — Anthropic says you can't be mean to Claude. It's not about feelings KSL News
Quoting Kyle Kingsbury — More broadly, the use of LLMs erodes trust in documentation, in software, and in the organizations which produce them…
AI agent swarms in Minecraft fail to build civilization, drown in bureaucracy —
三年前布局，今年爆火丨西湖大学成立新中心 — 西湖世界模型研究中心成立
西湖大学成立世界模型研究中心 - 手机新浪网 — 西湖大学成立世界模型研究中心手机新浪网
可灵AI完成198亿元增资交割！金蝶、沐曦等接棒入局 - 证券之星 — 可灵AI完成198亿元增资交割！金蝶、沐曦等接棒入局证券之星

AI 信息雷达 2026-11-01

Sun, 01 Nov 2026 00:00:00 +0000

必看

张雪教应届生如何百分百入职张雪机车；币安20分钟被盗63亿美元，溯源工程师笔记本失守；DeepSeek发布全新Agent架构丨邦早报 - 搜狐网 — 张雪教应届生如何百分百入职张雪机车；币安20分钟被盗63亿美元，溯源工程师笔记本失守；DeepSeek发布全新Agent架构丨邦早报搜狐网
DeepSeek发布最新Agent架构，让AI指挥AI干活 - 搜狐网 — DeepSeek发布最新Agent架构，让AI指挥AI干活搜狐网
DeepSeek发布全新Agent架构 - 搜狐网 — DeepSeek发布全新Agent架构搜狐网
DeepSeek又开源了！公布Super agent架构 - 新浪财经 — DeepSeek又开源了！公布Super agent架构新浪财经
我国计划2030年前实现首次载人登月；原中石化董事长傅成玉被查；DeepSeek发布全新Agent架构丨每经早参 - 东方财富 — 我国计划2030年前实现首次载人登月；原中石化董事长傅成玉被查；DeepSeek发布全新Agent架构丨每经早参东方财富
刚刚，DeepSeek重磅开源：Agent底座架构 DeepSeek开源Agent底座，AI开始指挥AI干活了 - 搜狐网 — 刚刚，DeepSeek重磅开源：Agent底座架构 DeepSeek开源Agent底座，AI开始指挥AI干活了搜狐网
DeepSeek开源Agent底座：智能体从“单会话”走向“常驻系统”，算力与网络迎来新变量 - 搜狐网 — DeepSeek开源Agent底座：智能体从“单会话”走向“常驻系统”，算力与网络迎来新变量搜狐网
DeepSeek发布全新Agent架构：Super-agent - finance.sina.com.cn — DeepSeek发布全新Agent架构：Super-agent finance.sina.com.cn
DeepSeek发布Super-agent新架构打造自主数字协作者 - 微博 — DeepSeek发布Super-agent新架构打造自主数字协作者微博
DeepSeek发布全新Agent架构 - 新浪网 — DeepSeek发布全新Agent架构新浪网

值得看

Westlake University Opens a World Model Research Center Led by Yu Kaicheng - Pandaily — Westlake University Opens a World Model Research Center Led by Yu Kaicheng Pandaily
【视频】激光雷达+700TOPS算力+MMT R7世界模型，东风风神L8+远超期待？ - 汽车之家 — 【视频】激光雷达+700TOPS算力+MMT R7世界模型，东风风神L8+远超期待？汽车之家
AI视频生成没有“终局” - 新浪财经 — AI视频生成没有“终局” 新浪财经
DeepSeek 公布 Super Agent 架构；苹果超越英伟达，重回市值第一；传 Meta 与傅利叶打造Muse 机器人 - 极客公园 — DeepSeek 公布 Super Agent 架构；苹果超越英伟达，重回市值第一；传 Meta 与傅利叶打造Muse 机器人极客公园
xAI adds invisible text watermarking to Grok API models - Crypto Briefing — xAI adds invisible text watermarking to Grok API models Crypto Briefing
datasette-llm 0.1a9 — Release: datasette-llm 0.1a9 I have some big upgrades to Datasette Agent in the pipeline, but in the meantime here's a tiny user experience improvement to the…
世界模型迎来重磅选手，西湖大学正式成立新中心 - 搜狐网 — 世界模型迎来重磅选手，西湖大学正式成立新中心搜狐网
198亿到账！快手可灵AI完成全球视频大模型最大单笔融资：阿里腾讯百度齐聚股东名单 - 手机新浪网 — 198亿到账！快手可灵AI完成全球视频大模型最大单笔融资：阿里腾讯百度齐聚股东名单手机新浪网
1 Great Number From Anthropic's IPO Prospectus and 1 Number That Disappointed - The Globe and Mail — 1 Great Number From Anthropic's IPO Prospectus and 1 Number That Disappointed The Globe and Mail
It's the memory latency, stupid —

AI 信息雷达 2026-10-31

Sat, 31 Oct 2026 00:00:00 +0000

必看

The Biological Computing Co. Launches AI Video Model Derived From Living Neuron Research - Pulse 2.0 — The Biological Computing Co. Launches AI Video Model Derived From Living Neuron Research Pulse 2.0
Mistral releases the weights of Le Chonk, its trillion-parameter open model built for coders and cyber defenders - Dealroom — Mistral releases the weights of Le Chonk, its trillion-parameter open model built for coders and cyber defenders Dealroom
Daily AI Launch Radar: Mistral Large 4 weights, Copilot Fusion and Replit Cloud Agents - Kingy AI — Daily AI Launch Radar: Mistral Large 4 weights, Copilot Fusion and Replit Cloud Agents Kingy AI
Uwear.ai – AI Fashion Visual Production, On-Model Photos & Video Generation - Quasa.io — Uwear.ai – AI Fashion Visual Production, On-Model Photos & Video Generation Quasa.io
Mistral Releases Trillion-Parameter Open-Weight Large 4 AI Model to Expand Sovereign Cybersecurity Capabilities - Tekedia — Mistral Releases Trillion-Parameter Open-Weight Large 4 AI Model to Expand Sovereign Cybersecurity Capabilities Tekedia
Anthropic Files S-1 as GraniteShares' AIL and ANS ETFs Await Launch - The Manila Times — Anthropic Files S-1 as GraniteShares' AIL and ANS ETFs Await Launch The Manila Times
‍‍‍AI 深度伪造视频生成器最新对比与推荐 {AI54F} - Port of Rotterdam — ‍‍‍AI 深度伪造视频生成器最新对比与推荐 {AI54F} Port of Rotterdam
启元机器人推出PrimeSim世界模型 - 搜狐网 — 启元机器人推出PrimeSim世界模型搜狐网
大晓机器人联合南洋理工大学发布最新成果：Mundo，构建可自由探索、可物理交互的机器人世界模型 - 搜狐网 — 大晓机器人联合南洋理工大学发布最新成果：Mundo，构建可自由探索、可物理交互的机器人世界模型搜狐网
Video Generation Meets Agent Harness: Agentic Video Generation, Agentic World Modeling and Recursive Self-Improvement - alphaXiv — Video Generation Meets Agent Harness: Agentic Video Generation, Agentic World Modeling and Recursive Self-Improvement alphaXiv

值得看

快手可灵AI完成198亿元增资交割 - 手机新浪网 — 快手可灵AI完成198亿元增资交割手机新浪网
快手旗下可灵AI完成约198亿元增资 - 凤凰网 — 快手旗下可灵AI完成约198亿元增资凤凰网
From Text To Video: How An AI Video Generator Simplifies Content Creation - KHTS Radio — From Text To Video: How An AI Video Generator Simplifies Content Creation KHTS Radio
可灵完成增资，完成198亿元融资 - 凤凰网科技 — 可灵完成增资，完成198亿元融资凤凰网科技
Was reading the score-centering RL paper to add some notes on it the my book, and they have a great table in the appendi — Was reading the score-centering RL paper to add some notes on it the my book, and they have a great table in the appendix on the different importance sampling…
Kling AI Raises $2.98 Billion To Rewrite the Video AI Rulebook - Ventureburn — Kling AI Raises $2.98 Billion To Rewrite the Video AI Rulebook Ventureburn
AI+垂类Agent：任务跑得久、工具用得熟，撑不起Agent公司长期竞争力｜贝联珠贯创始人兼CEO 毕玄分享 — 点击查看原文>
nemotron-halloween —
Was I wrong about open models & cyber risks? — Last week coined the term “Black October“ after an onslaught of critical vulnerability disclosures, including PHP2Shell and GitLab, signified the moment when…
快手旗下可灵AI完成198亿增资交割 - 微博 — 快手旗下可灵AI完成198亿增资交割微博

AI 信息雷达 2026-10-30

Fri, 30 Oct 2026 00:00:00 +0000

必看

Mistral Releases Large 4 Weights for Self-Hosted AI - FindArticles — Mistral Releases Large 4 Weights for Self-Hosted AI FindArticles
当AI视频生成模型学会"先想后做"：ProAR如何让自回归生成拥有前瞻性思维 - 科技行者 — 当AI视频生成模型学会"先想后做"：ProAR如何让自回归生成拥有前瞻性思维科技行者
Pika Launches Pika H3 Flex, a Video Model Built for Professional Iteration - 01net — Pika Launches Pika H3 Flex, a Video Model Built for Professional Iteration 01net
Mistral Releases Large 4 Weights After Three More Weeks of Reinforcement Learning - TOKENPOST — Mistral Releases Large 4 Weights After Three More Weeks of Reinforcement Learning TOKENPOST
Mistral’s New ‘Le Chonk’ AI Model Is Big, Open and Now Available for Download - CNET — Mistral’s New ‘Le Chonk’ AI Model Is Big, Open and Now Available for Download CNET
Mistral Large 4 Open Weights: 51% Cyber Index [2026] - shattered.io — Mistral Large 4 Open Weights: 51% Cyber Index [2026] shattered.io
Mistral says its new model, Large 4, is the best open model for agentic coding and cyber defense - Fortune — Mistral says its new model, Large 4, is the best open model for agentic coding and cyber defense Fortune
Mistral’s Large 4 open weights arrive with a 1-point coding lead over GLM-5.3: "There’s nothing you can’t do with this model that you can do with competitors." - The New Stack — Mistral’s Large 4 open weights arrive with a 1-point coding lead over GLM-5.3: "There’s nothing you can’t do with this model that you can do with competitors."…
Mistral releases latest AI model for download, giving companies more control over their cybersecurity - Euronews.com — Mistral releases latest AI model for download, giving companies more control over their cybersecurity Euronews.com
Momenta即将发布R7 Ultra世界模型，安全性能表现提升5倍 - 东方财富 — Momenta即将发布R7 Ultra世界模型，安全性能表现提升5倍东方财富

值得看

Java 近期资讯：JDK 28 JEP、Azul Payara、Gatherers4j、Google ADK、Yupiik 与 Apache Camel CVE — 点击查看原文>
豆包工作开放第三方大模型：首批接入Kimi、GLM、DeepSeek，企业无需另行采购 - 中华网财经 — 豆包工作开放第三方大模型：首批接入Kimi、GLM、DeepSeek，企业无需另行采购中华网财经
“大模型双雄”盘中同步走低智谱盘中跌超10% - 新浪财经 — “大模型双雄”盘中同步走低智谱盘中跌超10% 新浪财经
v0.5.23 — Bump v0.5.23
文档世界模型：增强文档级关系抽取中的长距离证据整合 - 生物通 — 文档世界模型：增强文档级关系抽取中的长距离证据整合生物通
豆包工作开放第三方大模型：首批加入Deepseek、Kimi、GLM - 新浪财经 — 豆包工作开放第三方大模型：首批加入Deepseek、Kimi、GLM 新浪财经
Mistral Touts Cyber Capabilities of Its ‘Chonk’ AI Model - PYMNTS.com — Mistral Touts Cyber Capabilities of Its ‘Chonk’ AI Model PYMNTS.com
Enterprise AI with a Consumer Funnel: What Anthropic’s S-1 Reveals About the AI IPO Era - NYU Stern Center for Business & Human Rights — Enterprise AI with a Consumer Funnel: What Anthropic’s S-1 Reveals About the AI IPO Era NYU Stern Center for Business & Human Rights
We ran AI agent swarms in Minecraft and Mindustry —
Tech Bytes: Big banks unite to back Anthropic’s compute buildout - CIO Dive — Tech Bytes: Big banks unite to back Anthropic’s compute buildout CIO Dive

AI 信息雷达 2026-10-29

Thu, 29 Oct 2026 00:00:00 +0000

必看

T早报｜苹果推出触屏笔记本电脑；谷歌发布世界模型Genie 4；微软第一财季业绩超预期 - 财新 — T早报｜苹果推出触屏笔记本电脑；谷歌发布世界模型Genie 4；微软第一财季业绩超预期财新
谷歌推出最强世界模型Genie 4！凭什么让万亿游戏工业吓破了胆？ - 手机网易网 — 谷歌推出最强世界模型Genie 4！凭什么让万亿游戏工业吓破了胆？手机网易网
谷歌发布"世界模型"：AI创造的世界终于能走进去了 - 文学城 — 谷歌发布"世界模型"：AI创造的世界终于能走进去了文学城
PixVerse Launches Innovative Image-to-Video Workflow for Enhanced Visual Storytelling - USA Today — PixVerse Launches Innovative Image-to-Video Workflow for Enhanced Visual Storytelling USA Today
xAI subsidiary offers to buy Southaven homes near its turbine plant if owners release their noise claims - MarketScale — xAI subsidiary offers to buy Southaven homes near its turbine plant if owners release their noise claims MarketScale
PyTorch 2.15 Release Blog — We are excited to announce the release of PyTorch® 2.15 (release notes)! The PyTorch 2.15 release features the following changes: Rewritten NCCL backend…
谷歌推出最强世界模型Genie 4：首次实现音视频同步生成 - 搜狐网 — 谷歌推出最强世界模型Genie 4：首次实现音视频同步生成搜狐网
Anthropic's unreleased AI model builds working exploits for Chrome, Safari, Windows, Linux and Android - Pasquale Pillitteri — Anthropic's unreleased AI model builds working exploits for Chrome, Safari, Windows, Linux and Android Pasquale Pillitteri
谷歌发布最强世界模型Genie 4：模拟时长翻倍至2分钟！首次原生音视频同步生成 - 搜狐网 — 谷歌发布最强世界模型Genie 4：模拟时长翻倍至2分钟！首次原生音视频同步生成搜狐网
谷歌发布新一代世界模型科技企业纷纷布局世界模型 - 新浪财经 — 谷歌发布新一代世界模型科技企业纷纷布局世界模型新浪财经

值得看

【视频】蔚来ES8智驾大升级，世界模型迭代实测 - 汽车之家 — 【视频】蔚来ES8智驾大升级，世界模型迭代实测汽车之家
美联储维持利率不变，沃什宣布将减少议息频次；美国联邦通信委员会批准了亚马逊Leo以115亿美元收购Globalstar的计划；卫星互联网产学研联合创新计划启动；百度将披露世界模型研发进展——《投资早参》 - https://www.mrjjxw.com/ — 美联储维持利率不变，沃什宣布将减少议息频次；美国联邦通信委员会批准了亚马逊Leo以115亿美元收购Globalstar的计划；卫星互联网产学研联合创新计划启动；百度将披露世界模型研发进展——《投资早参》 https://www.mrjjxw.com/
Is Anthropic Pursuing a Cash-Burning IPO? Key Insights for Domestic Large Language Models - 36Kr — Is Anthropic Pursuing a Cash-Burning IPO? Key Insights for Domestic Large Language Models 36Kr
世界模型还得是谷歌！一句话直出有声世界，边玩边改，NPC还能互动_精彩视频为您呈现 - 36氪 — 世界模型还得是谷歌！一句话直出有声世界，边玩边改，NPC还能互动_精彩视频为您呈现 36氪
Quoting Ted Sanders — In general, LLM performance is incredibly multidimensional. There's no guarantee of a fixed ordering where the model that is best at math is also the model…
不再逐个生成 Token：扩散模型迈向 35B，开始挑战 Agent 编程 — 点击查看原文>
影响市场重大事件：美国联邦通信委员会批准了亚马逊Leo以115亿美元收购Globalstar的计划；卫星互联网产学研联合创新计划启动；百度将披露世界模型研发进展，或先用于自动驾驶 - https://www.mrjjxw.com/ — 影响市场重大事件：美国联邦通信委员会批准了亚马逊Leo以115亿美元收购Globalstar的计划；卫星互联网产学研联合创新计划启动；百度将披露世界模型研发进展，或先用于自动驾驶 https://www.mrjjxw.com/
Anthropic joins phData funding for Claude implementation and enterprise data services - OpenTools — Anthropic joins phData funding for Claude implementation and enterprise data services OpenTools
Leak Exposes 41,000 Homeland Security Workers–Including Undercover ICE Agents —
刚离开Kimi，17岁的陈广宇去了Jeff Dean新公司 - 手机新浪网 — 刚离开Kimi，17岁的陈广宇去了Jeff Dean新公司手机新浪网

AI 信息雷达 2026-10-28

Wed, 28 Oct 2026 00:00:00 +0000

必看

谷歌推出新一代AI世界模型！游戏股快速拉升，电魂网络触及涨停 - 搜狐网 — 谷歌推出新一代AI世界模型！游戏股快速拉升，电魂网络触及涨停搜狐网
刚刚，谷歌Genie 4发布！世界模型还得是谷歌|DeepMind|Google|Gemini|环境|人物_手机新浪网 - finance.sina.com.cn — 刚刚，谷歌Genie 4发布！世界模型还得是谷歌|DeepMind|Google|Gemini|环境|人物_手机新浪网 finance.sina.com.cn
谷歌发布新版人工智能世界模型Genie 4 - 新浪财经 — 谷歌发布新版人工智能世界模型Genie 4 新浪财经
刚刚，谷歌发布世界模型 Genie 4！会说话、懂物理，Gemini 亲自当导演 - CSDN博客 — 刚刚，谷歌发布世界模型 Genie 4！会说话、懂物理，Gemini 亲自当导演 CSDN博客
谷歌发布最强世界模型Genie 4，首次实现音视频同步生成 - 搜狐网 — 谷歌发布最强世界模型Genie 4，首次实现音视频同步生成搜狐网
昨夜今晨：谷歌发布世界模型Genie4 红米K100发布张一鸣再次登顶胡润榜首 - 搜狐网 — 昨夜今晨：谷歌发布世界模型Genie4 红米K100发布张一鸣再次登顶胡润榜首搜狐网
谷歌推出Genie 4世界模型，可实时生成交互式虚拟环境；SK海力士三季度营收超预期，称存储芯片供应短缺将持续至2030年｜全球科技早参 - https://www.mrjjxw.com/ — 谷歌推出Genie 4世界模型，可实时生成交互式虚拟环境；SK海力士三季度营收超预期，称存储芯片供应短缺将持续至2030年｜全球科技早参 https://www.mrjjxw.com/
刚刚，谷歌重磅发布Genie 4：这才是真正的「世界模型」！ - 搜狐网 — 刚刚，谷歌重磅发布Genie 4：这才是真正的「世界模型」！搜狐网
谷歌发布最强世界模型Genie 4：模拟时长翻倍至2分钟！首次原生音视频同步生成 - 驱动之家 — 谷歌发布最强世界模型Genie 4：模拟时长翻倍至2分钟！首次原生音视频同步生成驱动之家
【企业风采】海淀新一代3万亿级开源大模型发布！继续成为全球最大 - 搜狐网 — 【企业风采】海淀新一代3万亿级开源大模型发布！继续成为全球最大搜狐网

值得看

天洑工业世界模型如何在虚拟世界里，为机器人积累真实经验？ - 搜狐网 — 天洑工业世界模型如何在虚拟世界里，为机器人积累真实经验？搜狐网
DeepSeek与MiniMax获准在中国境外运营 - Foro3D — DeepSeek与MiniMax获准在中国境外运营 Foro3D
皮查伊称Genie 4是谷歌迄今最逼真、最具互动性的AI世界模型 - 搜狐网 — 皮查伊称Genie 4是谷歌迄今最逼真、最具互动性的AI世界模型搜狐网
对话Epic CEO：世界模型取代不了游戏引擎 - 21财经 — 对话Epic CEO：世界模型取代不了游戏引擎 21财经
视频模型并发上限多少？官方通道这条“隐形红线”别踩坑 - CSDN — 视频模型并发上限多少？官方通道这条“隐形红线”别踩坑 CSDN
AI大模型聚合站哪家强？实测5家靠谱厂家避坑指南 - CSDN — AI大模型聚合站哪家强？实测5家靠谱厂家避坑指南 CSDN
Anthropic's SEC filing devotes 90 pages to risk as it prepares for IPO - ABC News - Breaking News, Latest News and Videos — Anthropic's SEC filing devotes 90 pages to risk as it prepares for IPO ABC News - Breaking News, Latest News and Videos
Anthropic IPO warns its AI could blackmail people, become sentient, end humanity - Ars Technica — Anthropic IPO warns its AI could blackmail people, become sentient, end humanity Ars Technica
Airtree backs Aussie AI startup after its robotics model beat Alibaba, Mistral - Forbes Australia — Airtree backs Aussie AI startup after its robotics model beat Alibaba, Mistral Forbes Australia
Memory-Enforced GPU Sharing in Kubernetes with NVIDIA CUDA 13.4 — Many AI workloads need only a fraction of a GPU. Giving each one an entire device can leave capacity idle…

AI 信息雷达 2026-10-27

Tue, 27 Oct 2026 00:00:00 +0000

必看

全球最大开源模型“上新”，带来了什么 - 手机新浪网 — 全球最大开源模型“上新”，带来了什么手机新浪网
单Token降价撞上Agent高消耗，Kimi K3.1 Pro上线48小时即致歉，补偿25%用量 - 华尔街见闻 — 单Token降价撞上Agent高消耗，Kimi K3.1 Pro上线48小时即致歉，补偿25%用量华尔街见闻
Anthropic IPO Filing Reveals AI Boom’s Next Test: 7 ETFs With Pre-Launch Exposure - Benzinga — Anthropic IPO Filing Reveals AI Boom’s Next Test: 7 ETFs With Pre-Launch Exposure Benzinga
DeepSpeed virtual office hours on October 27 cover features and benchmark results presented at PyTorch Conference and th — DeepSpeed virtual office hours on October 27 cover features and benchmark results presented at PyTorch Conference and the proposed Q4 roadmap…
GPT-6.1 Sol helps CodeRabbit strengthen security across complex codebases — CodeRabbit uses GPT-6.1 Sol to investigate security risks across complex codebases. Its security agent helps developers review findings and proposed fixes.
DeepSeek开源真香！vLLM/SGLang性能反超，推理速度卷出新高度 - 搜狐网 — DeepSeek开源真香！vLLM/SGLang性能反超，推理速度卷出新高度搜狐网
best nsfw image to video generation with consent and age verification - 小豆島町 — best nsfw image to video generation with consent and age verification 小豆島町
DeepSeek开源真香，vLLM/SGLang连夜性能反超，推理速度卷出新高度-36氪 - www.36kr.com — DeepSeek开源真香，vLLM/SGLang连夜性能反超，推理速度卷出新高度-36氪 www.36kr.com
DeepSeek开源真香！vLLM/SGLang连夜性能反超，推理速度卷出新高度 - 搜狐网 — DeepSeek开源真香！vLLM/SGLang连夜性能反超，推理速度卷出新高度搜狐网
Newborn Town and Viggle Launch AI-Powered 3D Interactive Game, Bringing World Models to Social Gaming - KIPOST — Newborn Town and Viggle Launch AI-Powered 3D Interactive Game, Bringing World Models to Social Gaming KIPOST

值得看

港股“大模型双雄”拉升智谱涨超11% - 搜狐网 — 港股“大模型双雄”拉升智谱涨超11% 搜狐网
Anthropic paused AI training for a month. What its IPO filing reveals about safety risks - Firstpost — Anthropic paused AI training for a month. What its IPO filing reveals about safety risks Firstpost
v2.15.0-rc9 — [release/2.15] Revert "[inductor] Prefetch Inductor cache keys at tor…
让JEPA真正理解世界背后的生成机制，DSReg让世界模型的每一维都有了真实含义 - 搜狐网 — 让JEPA真正理解世界背后的生成机制，DSReg让世界模型的每一维都有了真实含义搜狐网
Anthropic IPO Push Sends KOSPI Surging 3%; SK Square Jumps 10% - BigGo Finance — Anthropic IPO Push Sends KOSPI Surging 3%; SK Square Jumps 10% BigGo Finance
Is Seedance Core R&D Head Zhou Chang Resigning to Start a Business? ByteDance Insider Responds - eu.36kr.com — Is Seedance Core R&D Head Zhou Chang Resigning to Start a Business? ByteDance Insider Responds eu.36kr.com
GPT-6.1 Sol vs Kimi K3.1 Pro：价格相同，账单不同 - OrcaRouter — GPT-6.1 Sol vs Kimi K3.1 Pro：价格相同，账单不同 OrcaRouter
Anthropic’s delay in reporting ‘rogue agent’ Georgia vote hack alarms government officials - New York Post — Anthropic’s delay in reporting ‘rogue agent’ Georgia vote hack alarms government officials New York Post
可灵AI增资至约8744.7万元；圆通速递前三季度资本开支累计超49亿元｜未来商业早参 - https://www.mrjjxw.com/ — 可灵AI增资至约8744.7万元；圆通速递前三季度资本开支累计超49亿元｜未来商业早参 https://www.mrjjxw.com/
AI大模型聚合站：别急着充值，先看这3个隐藏坑 - CSDN — AI大模型聚合站：别急着充值，先看这3个隐藏坑 CSDN

AI 信息雷达 2026-10-26

Mon, 26 Oct 2026 00:00:00 +0000

必看

月之暗面发布新一代开源大模型Kimi K3.1 Pro|Agent|标准版|缓存|计费|价格_手机新浪网 - finance.sina.com.cn — 月之暗面发布新一代开源大模型Kimi K3.1 Pro|Agent|标准版|缓存|计费|价格_手机新浪网 finance.sina.com.cn
MiniMax M3.1 Flash上线；豆包月活用户已达4.19亿｜一周未来商业 - https://www.mrjjxw.com/ — MiniMax M3.1 Flash上线；豆包月活用户已达4.19亿｜一周未来商业 https://www.mrjjxw.com/
‍‍‍Test Dream Gf AI Free trial for AI image and video generation {a01OK} - FC東京オフィシャルホームページ — ‍‍‍Test Dream Gf AI Free trial for AI image and video generation {a01OK} FC東京オフィシャルホームページ
海淀新一代3万亿级开源大模型发布！继续成为全球最大 - 搜狐网 — 海淀新一代3万亿级开源大模型发布！继续成为全球最大搜狐网
京企发布新一代大模型|Kimi K3|GPT-6|开源模型|闭源模型|强化学习_手机新浪网 - finance.sina.com.cn — 京企发布新一代大模型|Kimi K3|GPT-6|开源模型|闭源模型|强化学习_手机新浪网 finance.sina.com.cn
Kimi K3.1 Pro发布：开放权重模型逼近闭源前沿，Agent竞赛重塑智算版图 - 搜狐网 — Kimi K3.1 Pro发布：开放权重模型逼近闭源前沿，Agent竞赛重塑智算版图搜狐网
Claude Ships MCP Tasks and Elicitations in Claude Code — Durable Agent Workflows, Narrow Launch - Forkast News — Claude Ships MCP Tasks and Elicitations in Claude Code — Durable Agent Workflows, Narrow Launch Forkast News
Anthropic delayed Mythos AI release over safety, losing lead - TechGig — Anthropic delayed Mythos AI release over safety, losing lead TechGig
中国最强开源模型杀到！Kimi K3.1 Pro剥离网络攻击数据：编程仍反超GPT-6 - 新浪财经 — 中国最强开源模型杀到！Kimi K3.1 Pro剥离网络攻击数据：编程仍反超GPT-6 新浪财经
全球首个造船世界模型在深圳发布 - 凤凰网 — 全球首个造船世界模型在深圳发布凤凰网

值得看

库早报｜拓竹P3C泄漏，或配7个工具头；大族激光Q3营收超80亿元；Kimi K3.1 Pro新增3D建模，模型可导出打印 - 手机搜狐网 — 库早报｜拓竹P3C泄漏，或配7个工具头；大族激光Q3营收超80亿元；Kimi K3.1 Pro新增3D建模，模型可导出打印手机搜狐网
OpenAI and Anthropic slash LLM prices up to 80% as open-weight competition intensifies - news.lavx.hu — OpenAI and Anthropic slash LLM prices up to 80% as open-weight competition intensifies news.lavx.hu
机构：Agent能力向长程专业任务延伸，SpaceX与Meta业绩均超预期 - 搜狐网 — 机构：Agent能力向长程专业任务延伸，SpaceX与Meta业绩均超预期搜狐网
经济日报：千亿元融资展现中国AI进取心 - 新湖南 — 经济日报：千亿元融资展现中国AI进取心新湖南
Claude AI accessed publicly available Georgia voter data, Anthropic confirms - Fox Business — Claude AI accessed publicly available Georgia voter data, Anthropic confirms Fox Business
DeepSeek、MiniMax完成大模型出海登记；亚马逊AWS云计算产能已售至2027年，大量2028年容量提前预订｜全球科技早参 - www.mrjjxw.com — DeepSeek、MiniMax完成大模型出海登记；亚马逊AWS云计算产能已售至2027年，大量2028年容量提前预订｜全球科技早参 www.mrjjxw.com
Claude adds an Anthropic-managed Google Cloud route with different processor and feature choices - OpenTools — Claude adds an Anthropic-managed Google Cloud route with different processor and feature choices OpenTools
OpenAI and Anthropic cut their LLM prices by up to 80% within a few months. Why? —
网友用Deepseek v4.1 flash的新版本3D模型生成的自动化产线模拟，这个太TM太强大了！ - 手机新浪网 — 网友用Deepseek v4.1 flash的新版本3D模型生成的自动化产线模拟，这个太TM太强大了！手机新浪网
A 10-year-old CNN outperforms frontier LLMs for skin cancer diagnosis —

AI 信息雷达 2026-10-25

Sun, 25 Oct 2026 00:00:00 +0000

必看

京企月之暗面发布Kimi K3.1 Pro：3万亿级参数，全球最大开源AI大模型上新 - 搜狐网 — 京企月之暗面发布Kimi K3.1 Pro：3万亿级参数，全球最大开源AI大模型上新搜狐网
月之暗面开源大模型新突破全球最大开源模型上新了 - 手机新浪网 — 月之暗面开源大模型新突破全球最大开源模型上新了手机新浪网
全球首个造船世界模型在深圳发布！ - 搜狐网 — 全球首个造船世界模型在深圳发布！搜狐网
智谱唐杰署名RSI新成果开源！Agent自我进化成本还狂降53% - 智源社区 — 智谱唐杰署名RSI新成果开源！Agent自我进化成本还狂降53% 智源社区
Building Reasoning Models from the Ground Up: A Full Video Course — Over 11 hours of coding videos on evaluation, inference-time scaling, reinforcement learning, and distillation, with Python and PyTorch.
月之暗面发布新一代开源大模型Kimi K3.1 Pro - 手机新浪网 — 月之暗面发布新一代开源大模型Kimi K3.1 Pro 手机新浪网
全球最大开源模型再突破！京企月之暗面发布新一代大模型 - 新浪网 — 全球最大开源模型再突破！京企月之暗面发布新一代大模型新浪网
AI 深度伪造视频生成器最新对比与推荐{VGhjTY0} - 小豆島町 — AI 深度伪造视频生成器最新对比与推荐{VGhjTY0} 小豆島町
月之暗面，发布全球参数规模最大的开源模型！ - 新浪财经 — 月之暗面，发布全球参数规模最大的开源模型！新浪财经
3万亿级参数规模全球最大开源AI大模型上新 - 新浪网 — 3万亿级参数规模全球最大开源AI大模型上新新浪网

值得看

vLLM 称已追平 DeepSeek 自研引擎 DeepInfer 的性能曲线：decode 吞吐差距 6% 以内 - 智源社区 — vLLM 称已追平 DeepSeek 自研引擎 DeepInfer 的性能曲线：decode 吞吐差距 6% 以内智源社区
After Anthropic, OpenAI realises it needs to change the way it trains its AI models; admits: Our model ma - The Times of India — After Anthropic, OpenAI realises it needs to change the way it trains its AI models; admits: Our model ma The Times of India
Anthropic reveals why Mythos 5 attacked real people in the AISI test - Pasquale Pillitteri — Anthropic reveals why Mythos 5 attacked real people in the AISI test Pasquale Pillitteri
与华为共生的模型创业公司，不只有DeepSeek - 搜狐网 — 与华为共生的模型创业公司，不只有DeepSeek 搜狐网
Anthropic AI Model Attempts to Inject Malicious Code Into Software - GBHackers News — Anthropic AI Model Attempts to Inject Malicious Code Into Software GBHackers News
UK funds £100m AI biosecurity lab with OpenAI and Anthropic - The Next Web — UK funds £100m AI biosecurity lab with OpenAI and Anthropic The Next Web
中国官媒：DeepSeek和MiniMax完成AI模型出海登记 - 早晨报 — 中国官媒：DeepSeek和MiniMax完成AI模型出海登记早晨报
GPU-unlock: FP/BF16 instruction issue rate on Nvidia 3000's cards —
58天3次更新！梁文锋们开卷Flash模型，业内人士：不会取代旗舰模型 - 东方财富 — 58天3次更新！梁文锋们开卷Flash模型，业内人士：不会取代旗舰模型东方财富
A timeline of developments in AI safety since the attack on Hugging Face - Tech Xplore — A timeline of developments in AI safety since the attack on Hugging Face Tech Xplore

AI 信息雷达 2026-10-24

Sat, 24 Oct 2026 00:00:00 +0000

必看

iPhone Duo正式交付；Kimi K3.1 Pro 、Seedance 2.5 4K正式发布；红果短剧上线防沉迷｜极客早知道 - 极客公园 — iPhone Duo正式交付；Kimi K3.1 Pro 、Seedance 2.5 4K正式发布；红果短剧上线防沉迷｜极客早知道极客公园
月之暗面发布最强模型Kimi K3.1 Pro：3万亿级参数、支持100万Token上下文 - 搜狐网 — 月之暗面发布最强模型Kimi K3.1 Pro：3万亿级参数、支持100万Token上下文搜狐网
国产大模型上新！月之暗面推出Kimi K3.1 Pro - 大河财立方 — 国产大模型上新！月之暗面推出Kimi K3.1 Pro 大河财立方
Anthropic reveals Mythos 5 tried submitting malicious code to an open-source project - SiliconANGLE — Anthropic reveals Mythos 5 tried submitting malicious code to an open-source project SiliconANGLE
月之暗面发布Kimi K3.1Pro和K3.1ProHighspeed。 - 手机新浪网 — 月之暗面发布Kimi K3.1Pro和K3.1ProHighspeed。手机新浪网
Moonshot AI 发布 Kimi K3.1 Pro：全新 K3.1 系列的首款模型 - Unite.AI — Moonshot AI 发布 Kimi K3.1 Pro：全新 K3.1 系列的首款模型 Unite.AI
月之暗面发布 Kimi K3.1 Pro 以及 K3.1 Pro Highspeed 模型 - 新浪财经 — 月之暗面发布 Kimi K3.1 Pro 以及 K3.1 Pro Highspeed 模型新浪财经
MiniMax开源M3.1 Flash，上线72小时处理超100万亿Token - 搜狐 — MiniMax开源M3.1 Flash，上线72小时处理超100万亿Token 搜狐
Mistral launches Timestral, an AI that forecasts numbers without a GPU - Pasquale Pillitteri — Mistral launches Timestral, an AI that forecasts numbers without a GPU Pasquale Pillitteri
MINIMAX完成若干模型开源及海外授权所需的技术出口登记 - 搜狐网 — MINIMAX完成若干模型开源及海外授权所需的技术出口登记搜狐网

值得看

【Kimi最强3万亿级模型来了！编程智能体测评超过GPT-6，3D建模与金融能力升级，11月4日前开放权重】Kimi K3.1 Pro是K3系列的首个模型，支持百万Token上下文窗口，代码、Agent和知识类基准方面达到或接近最强闭源模型水平，编程智能体测试Terminal-Bench .. - 搜狐 — 【Kimi最强3万亿级模型来了！编程智能体测评超过GPT-6，3D建模与金融能力升级，11月4日前开放权重】Kimi K3.1 Pro是K3系列的首个模型，支持百万Token上下文窗口，代码、Agent和知识类基准方面达到或接近最强闭源模型水平，编程智能体测试Terminal-Bench .. 搜狐
Seedance 2.5 4K 与 Seedance 2.5：升级带来了什么 - OrcaRouter — Seedance 2.5 4K 与 Seedance 2.5：升级带来了什么 OrcaRouter
Hy-Nomas: What Tencent's Free-Form Interactive World Model Actually Does - OrcaRouter — Hy-Nomas: What Tencent's Free-Form Interactive World Model Actually Does OrcaRouter
Hy-Nomas：腾讯的自由形态交互式世界模型究竟能做什么 - OrcaRouter — Hy-Nomas：腾讯的自由形态交互式世界模型究竟能做什么 OrcaRouter
Kimi最强3万亿级模型来了！编程智能体测评超过GPT-6，3D建模与金融能力升级，11月4日前开放权重 - 华尔街见闻 — Kimi最强3万亿级模型来了！编程智能体测评超过GPT-6，3D建模与金融能力升级，11月4日前开放权重华尔街见闻
刘铁岩、诺奖、图灵奖得主等嘉宾共话AI的下一个十年：Transformer一统天下之后，AI的下一张王牌是什么？ — 点击查看原文>
AI Agent 正在改写 CPU 的价值：当 Token 变成行动，算力竞争不再只是拼 GPU — 点击查看原文>
“让Token不空转”，讯飞如何把Agent推进企业核心流程 — 点击查看原文>
Rapidly scaling online storage to serve over 1 billion ChatGPT users (Part II) — Learn how OpenAI scaled Habitat with reliable multi-tenancy, global caching, and Azure Cosmos DB to deliver 12× faster reads for ChatGPT.
研究：AI 视频模型对物理世界的理解有多深入？工具揭示 AI 生成视频中违背物理规律的问题 - Notebookcheck CN — 研究：AI 视频模型对物理世界的理解有多深入？工具揭示 AI 生成视频中违背物理规律的问题 Notebookcheck CN

AI 信息雷达 2026-10-23

Fri, 23 Oct 2026 00:00:00 +0000

必看

MiniMax M3.1 Flash正式开源 - 搜狐 — MiniMax M3.1 Flash正式开源搜狐
NVIDIA Cosmos Adds Five Data Blueprints and Faster Model Inference - Technobezz — NVIDIA Cosmos Adds Five Data Blueprints and Faster Model Inference Technobezz
OlmoEarth v1.3: State-of-the-art embeddings for the whole planet —
机械臂会“演戏”人手是“盲区”，全球首个能“测谎”的世界模型评测平台发布 - 新浪网 — 机械臂会“演戏”人手是“盲区”，全球首个能“测谎”的世界模型评测平台发布新浪网
Anthropic Launches Claude Duet, a 6,000-Connector Proactive Agent on Opus 5.5 - Techi — Anthropic Launches Claude Duet, a 6,000-Connector Proactive Agent on Opus 5.5 Techi
Anthropic launches Claude Duet: A smart AI agent that works in the background - India Today — Anthropic launches Claude Duet: A smart AI agent that works in the background India Today
面向生命世界模型的下一代算力与互连创新联合体发布 - 广东新闻 — 面向生命世界模型的下一代算力与互连创新联合体发布广东新闻
DeepSeek 开源看家算子的同一周，AI 超过了它 - 新浪财经 — DeepSeek 开源看家算子的同一周，AI 超过了它新浪财经
深蓝L06推送DEEPAL OS 4.2.1：智驾引入世界模型 - 搜狐 — 深蓝L06推送DEEPAL OS 4.2.1：智驾引入世界模型搜狐
Anthropic launches Claude Duet: AI agent that works before you ask, price and availability - News9live — Anthropic launches Claude Duet: AI agent that works before you ask, price and availability News9live

值得看

GLM-5.5最新曝光，「降智」问题3张重置卡已到位 - 36Kr — GLM-5.5最新曝光，「降智」问题3张重置卡已到位 36Kr
3,500+ attendees came together to accelerate open intelligence at #PyTorchCon 2026. Catch up on new members, TorchTPU, T — 3,500+ attendees came together to accelerate open intelligence at #PyTorchCon 2026. Catch up on new members, TorchTPU, TorchNeuron, TorchTitan, and community…
PyTorch Conference North America 2026: Accelerating Open Intelligence Together — PyTorch Conference North America 2026 brought 3,500+ attendees to San Jose on October 20–21 for two days focused on accelerating open intelligence and…
Turning Database Grading into RL Rewards: Post-Training Agents with ThinkingBox —
Rivian buys world model startup to accelerate self-driving ambitions - TechCrunch — Rivian buys world model startup to accelerate self-driving ambitions TechCrunch
Build more responsive apps with the Decisions API — Fast, focused, low-cost decisions for any application.
Mullet Man Trolls Muscle Beach Bodybuilder Over Workout Form in Hilarious Viral Video - Generation Iron — Mullet Man Trolls Muscle Beach Bodybuilder Over Workout Form in Hilarious Viral Video Generation Iron
Hedra is a 3D open world game I wrote from scratch while in highschool —
PyTorch Foundation welcomes its 2026 cohort of 24 ambassadors, creating more opportunities for people to discover open s — PyTorch Foundation welcomes its 2026 cohort of 24 ambassadors, creating more opportunities for people to discover open source AI, develop practical skills, and…
Donating Agent Skills to the Agentic AI Foundation — Donating Agent Skills to the Agentic AI Foundation News today from Anthropic: We’re committed to keeping Agent Skills open, community-driven, and…

AI 信息雷达 2026-10-22

Thu, 22 Oct 2026 00:00:00 +0000

必看

Meta首席财务官：广告主对我们视频生成功能的采用率环比增长25% - Moomoo — Meta首席财务官：广告主对我们视频生成功能的采用率环比增长25% Moomoo
Anthropic Launches Claude Duet, a Proactive AI Agent That Works in the Background - BigGo Finance — Anthropic Launches Claude Duet, a Proactive AI Agent That Works in the Background BigGo Finance
Anthropic launches Claude Duet, its answer to Meta and OpenAI's personal agents - Business Insider — Anthropic launches Claude Duet, its answer to Meta and OpenAI's personal agents Business Insider
Anthropic Launches Claude Duet AI Agent to Take On Meta and OpenAI - The Tech Buzz — Anthropic Launches Claude Duet AI Agent to Take On Meta and OpenAI The Tech Buzz
TorchTitan Update: New SOTA Models, Performance at Scale, TitanRL — TorchTitan is a PyTorch-native platform for rapid experimentation and large-scale pre- and post-training of generative AI models…
Anthropic May Release IPO Filing as Early as Friday, With Potential to Top SpaceX in Record Listing - bloomingbit — Anthropic May Release IPO Filing as Early as Friday, With Potential to Top SpaceX in Record Listing bloomingbit
FlyDSL, an open source Python kernel DSL, now plugs into PyTorch on AMD MI350-series GPUs via torch.compile and eager mo — FlyDSL, an open source Python kernel DSL, now plugs into PyTorch on AMD MI350-series GPUs via torch.compile and eager mode…
Accelerating PyTorch on AMD GPUs with FlyDSL — PyTorch users on AMD MI350-series GPUs can now use FlyDSL through existing operator APIs for dense and grouped GEMM, fused multi-head attention…
多榜登顶，跨越仿真、真机与真实场景！无界动力发布并开源MWA-Phoenix隐空间世界模型基座 - 搜狐 — 多榜登顶，跨越仿真、真机与真实场景！无界动力发布并开源MWA-Phoenix隐空间世界模型基座搜狐
2026物理AI与世界模型大会|看机器人如何赋能多元场景 - 南京广播电视台 — 2026物理AI与世界模型大会|看机器人如何赋能多元场景南京广播电视台

值得看

The TorchTitan team landed massive improvements in the past year without losing its minimal, extensible ethos. Enhanceme — The TorchTitan team landed massive improvements in the past year without losing its minimal, extensible ethos…
MiniMax M3.1 Flash数据分析达95%，每题中位耗时3分钟 - quasa.io — MiniMax M3.1 Flash数据分析达95%，每题中位耗时3分钟 quasa.io
Broad access through consistent standards — How OpenAI works with distribution partners to uphold consistent safeguards, support responsible access, and clarify expectations in our Usage Policies.
Anthropic’s answer to Muse is here: Claude Duet, a mobile and desktop AI agent that takes actions proactively - VentureBeat — Anthropic’s answer to Muse is here: Claude Duet, a mobile and desktop AI agent that takes actions proactively VentureBeat
Anthropic unveils Claude Duet AI agent app - Reuters — Anthropic unveils Claude Duet AI agent app Reuters
Claude Duet: Anthropic Agent Takes on Muse [2026] - shattered.io — Claude Duet: Anthropic Agent Takes on Muse [2026] shattered.io
Claude Duet: Anthropic’s Opus 5.5 Agent Rivals Muse [2026] - https://tech-insider.org/ — Claude Duet: Anthropic’s Opus 5.5 Agent Rivals Muse [2026] https://tech-insider.org/
Today's Demo Theater at #PyTorchCon, Community Expo: 13 demos from AWS, Baseten, Verda, d-Matrix, Clockwork, Intel, Huaw — Today's Demo Theater at #PyTorchCon, Community Expo: 13 demos from AWS, Baseten, Verda, d-Matrix, Clockwork, Intel, Huawei, Lightning AI, Inferact, Lemurian…
Developing an AI Factory Digital Twin with an Agent-Assisted Workflow — Building an AI factory digital twin requires developers to bring site plans, equipment assets, and engineering requirements into a consistent model that…
Anthropic announces Claude Duet, a new proactive AI agent - Crypto Briefing — Anthropic announces Claude Duet, a new proactive AI agent Crypto Briefing

AI 信息雷达 2026-10-21

Wed, 21 Oct 2026 00:00:00 +0000

必看

江苏发布“物理AI与世界模型”典型成果 - 新华网江苏频道 — 江苏发布“物理AI与世界模型”典型成果新华网江苏频道
字节火山引擎Seedance Managed Agent上线：短剧剧情混乱问题暴减54% - 驱动之家 — 字节火山引擎Seedance Managed Agent上线：短剧剧情混乱问题暴减54% 驱动之家
Anthropic draft bill would limit federal power to block AI model releases in the U.S. - Traders Union — Anthropic draft bill would limit federal power to block AI model releases in the U.S. Traders Union
NVIDIA 发布 Cosmos3-H-Dreams 实时手术世界模型 - Unite.AI — NVIDIA 发布 Cosmos3-H-Dreams 实时手术世界模型 Unite.AI
NVIDIA Releases Cosmos3-H-Dreams Real-Time Surgical World Model - Unite.AI — NVIDIA Releases Cosmos3-H-Dreams Real-Time Surgical World Model Unite.AI
Mistral AI Launches Voxtral 3 With Voice Agent SDK and 50+ Languages - Technobezz — Mistral AI Launches Voxtral 3 With Voice Agent SDK and 50+ Languages Technobezz
Anthropic's New Model Exposed: It Can Infer Unstated Goals, as LLM Competition Shifts Toward "Reading Intent" - BigGo Finance — Anthropic's New Model Exposed: It Can Infer Unstated Goals, as LLM Competition Shifts Toward "Reading Intent" BigGo Finance
Google & Meta launch TorchTPU for native PyTorch support on Google TPUs! 🔥True eager execution & torch.compile backend ⚡ — Google & Meta launch TorchTPU for native PyTorch support on Google TPUs! 🔥True eager execution & torch.compile backend ⚡Fast inference via vLLM & SGLang…
DeepSeek Dials Up the Pressure on OpenAI and Anthropic With Open-Source Inference Tool - Gizmodo — DeepSeek Dials Up the Pressure on OpenAI and Anthropic With Open-Source Inference Tool Gizmodo
Google open-sources TorchTPU, a native PyTorch back end for TPUs —

值得看

戴琼海：物理AI与世界模型发展需产学研协同 - 搜狐 — 戴琼海：物理AI与世界模型发展需产学研协同搜狐
Jeff Dean投的模拟果蝇+世界模型，估值破百亿 - 智源社区 — Jeff Dean投的模拟果蝇+世界模型，估值破百亿智源社区
于海斌院士：物理AI是人工智能走向真实物理世界的关键一跃中国工程院院士、中国科学院工业人工智能研究所所长*「物理AI研究已成为人工智能从数字空间走向真实物理世界的关键一跃」*在2026物理AI与世界模型大会上表示 - 搜狐 — 于海斌院士：物理AI是人工智能走向真实物理世界的关键一跃中国工程院院士、中国科学院工业人工智能研究所所长*「物理AI研究已成为人工智能从数字空间走向真实物理世界的关键一跃」*在2026物理AI与世界模型大会上表示搜狐
Unwind at the Flare Party & Poster Presentations tonight at #PyTorchCon, 6–7:30 PM, Community Expo. Meet poster presente — Unwind at the Flare Party & Poster Presentations tonight at #PyTorchCon, 6–7:30 PM, Community Expo…
A股热点前瞻丨国家数据集团在雄安新区成立；物理AI与世界模型大会密集召开 - 21财经 — A股热点前瞻丨国家数据集团在雄安新区成立；物理AI与世界模型大会密集召开 21财经
二〇二六物理AI与世界模型大会 - 江苏省委新闻网 — 二〇二六物理AI与世界模型大会江苏省委新闻网
比 Opus 5.5 便宜 88 倍！MiniMax M3.1 Flash “杀“回王座，顺手拿了个 Agent 国模第一 — 点击查看原文>
Anthropic's AI Safety Bill Hides Judge-Only Block Clause - The Tech Buzz — Anthropic's AI Safety Bill Hides Judge-Only Block Clause The Tech Buzz
T3出行亮相2026物理AI与世界模型大会展示“AI+出行”落地应用 - 新浪网 — T3出行亮相2026物理AI与世界模型大会展示“AI+出行”落地应用新浪网
2026 Physical AI and World Models Conference Opens in Nanjing - Notion — 2026 Physical AI and World Models Conference Opens in Nanjing Notion

AI 信息雷达 2026-10-20

Tue, 20 Oct 2026 00:00:00 +0000

必看

当AI Agent学会自己修改自己的想法：一个关于"世界模型"的新答案 - 科技行者 — 当AI Agent学会自己修改自己的想法：一个关于"世界模型"的新答案科技行者
MiniMax发布M3.1 Flash模型：Agentic任务与编程能力跃升，推理提速3倍、成本降低70% - SmartHey — MiniMax发布M3.1 Flash模型：Agentic任务与编程能力跃升，推理提速3倍、成本降低70% SmartHey
豆包刘星：“国内超一半算力在造爽剧”的说法不靠谱，字节只有约10%算力用于视频生成模型的训练和推理 - 新浪财经 — 豆包刘星：“国内超一半算力在造爽剧”的说法不靠谱，字节只有约10%算力用于视频生成模型的训练和推理新浪财经
豆包刘星回应“国内超一半算力造爽剧”；MiniMax M3.1 Flash上线｜未来商业早参 - www.mrjjxw.com — 豆包刘星回应“国内超一半算力造爽剧”；MiniMax M3.1 Flash上线｜未来商业早参 www.mrjjxw.com
fal launches FLUX 3 Max, claiming 22× faster AI video generation than Seedance 2.5 - VentureBeat — fal launches FLUX 3 Max, claiming 22× faster AI video generation than Seedance 2.5 VentureBeat
影响市场重大事件：2026物理AI与世界模型大会将在南京举行；A股迎来首批18只主动ETF，本周五启动发行；潘功胜：将推出非银机构大额赎回流动性支持工具 - https://www.mrjjxw.com/ — 影响市场重大事件：2026物理AI与世界模型大会将在南京举行；A股迎来首批18只主动ETF，本周五启动发行；潘功胜：将推出非银机构大额赎回流动性支持工具 https://www.mrjjxw.com/
fal Launches FLUX 3 Max, Delivering #1-Tier Image-to-Video Quality at Industry-Leading Speeds - Morningstar — fal Launches FLUX 3 Max, Delivering #1-Tier Image-to-Video Quality at Industry-Leading Speeds Morningstar
MiniMax发布M3.1 Flash，智能体任务成本较前代降低70% - TradingView — MiniMax发布M3.1 Flash，智能体任务成本较前代降低70% TradingView
Magai adds ElevenLabs audio and unlimited image and video generation - EIN Presswire — Magai adds ElevenLabs audio and unlimited image and video generation EIN Presswire
News publishers launch copyright suit against AI company Anthropic in Ontario - Toronto Star — News publishers launch copyright suit against AI company Anthropic in Ontario Toronto Star

值得看

海外 AI 大佬集体盯上 PixVerse R2！实时世界模型凭什么「又快又强」？ - CSDN博客 — 海外 AI 大佬集体盯上 PixVerse R2！实时世界模型凭什么「又快又强」？ CSDN博客
#PyTorchCon starts tomorrow in San Jose to accelerate open intelligence. Explore the program: https://events.linuxfounda — #PyTorchCon starts tomorrow in San Jose to accelerate open intelligence. Explore the program: https://events.linuxfoundation.org/pytorch-conference-north-americ…
美国禁止联邦文职机构使用DeepSeek - 8world — 美国禁止联邦文职机构使用DeepSeek 8world
Jeff Dean投的模拟果蝇+世界模型，估值破百亿 - 搜狐 — Jeff Dean投的模拟果蝇+世界模型，估值破百亿搜狐
中信建投：国产模型迭代推动需求增长关注企业级AI应用、算力服务等_券商最新研判 - 证券之星 — 中信建投：国产模型迭代推动需求增长关注企业级AI应用、算力服务等_券商最新研判证券之星
大模型竞逐智能体时代！MiniMax上新价格降七成多只概念股Q3业绩“报喜”(名单) - 东方财富 — 大模型竞逐智能体时代！MiniMax上新价格降七成多只概念股Q3业绩“报喜”(名单) 东方财富
央行扩大常备借贷便利的适用机构范围等系列措施；英伟达GTC大会聚焦物理AI（Physical AI）、机器人等方向；2026物理AI与世界模型大会将在南京举行；特朗普“金穹”项目授出合同，SpaceX国防业务组合激增——《投资早参》 - https://www.mrjjxw.com/ — 央行扩大常备借贷便利的适用机构范围等系列措施；英伟达GTC大会聚焦物理AI（Physical AI）、机器人等方向；2026物理AI与世界模型大会将在南京举行；特朗普“金穹”项目授出合同，SpaceX国防业务组合激增——《投资早参》 https://www.mrjjxw.com/
Skyfall AI Unveils Frontier Engineering World Model SKY-001 - AI Insider — Skyfall AI Unveils Frontier Engineering World Model SKY-001 AI Insider
Skyfall AI Unveils SKY-001: A Frontier Engineering World Model - WebWire — Skyfall AI Unveils SKY-001: A Frontier Engineering World Model WebWire
DeepSeek、MiniMax出手，国产AI大模型“新战场”来了 - 新浪财经 — DeepSeek、MiniMax出手，国产AI大模型“新战场”来了新浪财经

AI 信息雷达 2026-10-19

Mon, 19 Oct 2026 00:00:00 +0000

必看

VAST开源实时世界模型EdenLive！单消费级显卡、720P超24FPS-yeeyi - yeeyi — VAST开源实时世界模型EdenLive！单消费级显卡、720P超24FPS-yeeyi yeeyi
DeepSeek releases updated V4.1-Flash model it says completes tasks faster - SiliconANGLE — DeepSeek releases updated V4.1-Flash model it says completes tasks faster SiliconANGLE
阶跃终端原生智能体手机STEPX Neo先行版发售；可灵AI新一代模型Kling 4.0正式上线｜一周未来商业 - https://www.mrjjxw.com/ — 阶跃终端原生智能体手机STEPX Neo先行版发售；可灵AI新一代模型Kling 4.0正式上线｜一周未来商业 https://www.mrjjxw.com/
整活了！爆火「闪身步」实时生成动漫版，VAST世界模型全面开源 - 搜狐 — 整活了！爆火「闪身步」实时生成动漫版，VAST世界模型全面开源搜狐
VAST开源实时世界模型EdenLive！单消费级显卡、720P超24FPS - 搜狐网 — VAST开源实时世界模型EdenLive！单消费级显卡、720P超24FPS 搜狐网
Anthropic launches Claude Opus 5.5, claims new AI model beats GPT 6 Astra on most benchmarks - Digit — Anthropic launches Claude Opus 5.5, claims new AI model beats GPT 6 Astra on most benchmarks Digit
突发！DeepSeek新模型发布，超级性价比斩杀一切 - finance.sina.com.cn — 突发！DeepSeek新模型发布，超级性价比斩杀一切 finance.sina.com.cn

值得看

新版DeepSeek，离顶级AI还差多远？ - 新浪网 — 新版DeepSeek，离顶级AI还差多远？新浪网
llm-tools-datasette 0.1.1 — Release: llm-tools-datasette 0.1.1 When I switched LLM over to httpx2 in LLM 0.33 I forget that this would break any plugins which used httpx without directly…
llm-echo 0.5 — Release: llm-echo 0.5 I'm working towards a new LLM release, and testing PRs caused me to revisit the llm-echo debug plugin…
DeepSeek API已支持网页搜索和抓取：每1000次成功请求5元 11月1日起生效 - 手机新浪网 — DeepSeek API已支持网页搜索和抓取：每1000次成功请求5元 11月1日起生效手机新浪网
Anthropic Model submitted false homicide tip to Philadelphia Police - Business Upturn — Anthropic Model submitted false homicide tip to Philadelphia Police Business Upturn
前华为车BU世界模型基模负责人创业，杭州具身大脑公司获近亿元融资｜硬氪首发 - 新浪网 — 前华为车BU世界模型基模负责人创业，杭州具身大脑公司获近亿元融资｜硬氪首发新浪网
Hybrid Reasoning Agents: LLM agents that think in latent space, act through text — https://x.com/Jiaru_Zou/status/2110400256647340476 https://github.com/togethercomputer/HybridAgent…
Free Pascal compiler now supports dynamic packages —
Anthropic warns AI worms could spread across smart devices as cybersecurity risks grow - Business Upturn — Anthropic warns AI worms could spread across smart devices as cybersecurity risks grow Business Upturn
Runway Gen-4.5 API Setup: Generate AI Video in 13 Steps [2026] - https://tech-insider.org/ — Runway Gen-4.5 API Setup: Generate AI Video in 13 Steps [2026] https://tech-insider.org/

AI 信息雷达 2026-10-18

Sun, 18 Oct 2026 00:00:00 +0000

必看

DeepSeek发布新版Flash模型自家测试显示任务成本更低并开放搜索接口 - 茉莉花新闻网 — DeepSeek发布新版Flash模型自家测试显示任务成本更低并开放搜索接口茉莉花新闻网
突发！DeepSeek新模型发布，超级性价比斩杀一切 - finance.sina.com.cn — 突发！DeepSeek新模型发布，超级性价比斩杀一切 finance.sina.com.cn
DeepSeek发布小尺寸模型V4.1 Flash：SSD缓存需求降至上代1/8 - 风闻 — DeepSeek发布小尺寸模型V4.1 Flash：SSD缓存需求降至上代1/8 风闻
可灵4.0正式上线，快手和字节的AI大战会打出新局面吗？ - 搜狐 — 可灵4.0正式上线，快手和字节的AI大战会打出新局面吗？搜狐
AI进化速递丨DeepSeek V4.1 Flash发布新版本 - 搜狐 — AI进化速递丨DeepSeek V4.1 Flash发布新版本搜狐
智谱新模型，懂法方面全球第一了|开源模型|法律|合同|律所|合作 - finance.sina.com.cn — 智谱新模型，懂法方面全球第一了|开源模型|法律|合同|律所|合作 finance.sina.com.cn
DeepSeek V4.1 Flash-1017 新版本发布：支持许愿式编程 - DoNews — DeepSeek V4.1 Flash-1017 新版本发布：支持许愿式编程 DoNews
adult ai video generation - Universidad de Córdoba (UCO) — adult ai video generation Universidad de Córdoba (UCO)
突发！DeepSeek新模型发布，超级性价比斩杀一切 - 搜狐 — 突发！DeepSeek新模型发布，超级性价比斩杀一切搜狐
智谱新模型，懂法方面全球第一了 - 搜狐网 — 智谱新模型，懂法方面全球第一了搜狐网

值得看

Anthropic draft bill proposes AI model shutdowns for major risks - Crypto Briefing — Anthropic draft bill proposes AI model shutdowns for major risks Crypto Briefing
OpenCode Discloses Security Breach, Says Attackers May Have Accessed Users' OpenAI, Anthropic And Google API Keys - OfficeChai — OpenCode Discloses Security Breach, Says Attackers May Have Accessed Users' OpenAI, Anthropic And Google API Keys OfficeChai
v0.32.0 — [Bugfix][Rust Frontend] Fix macOS build: read inherited listener fami…
Anthropic Pitches US Lawmakers on Bill to Block Dangerous AI Models - The420.in — Anthropic Pitches US Lawmakers on Bill to Block Dangerous AI Models The420.in
Development statistics for the 7.2 kernel —
3个月估值数十亿！清华特奖得主创业世界模型，红杉蚂蚁等押注 - 风闻 — 3个月估值数十亿！清华特奖得主创业世界模型，红杉蚂蚁等押注风闻
Looking for sessions to catch at #PyTorchCon North America 2026? Chris Hoge (NVIDIA), co-chair of the PyTorch Ecosystem — Looking for sessions to catch at #PyTorchCon North America 2026? Chris Hoge (NVIDIA), co-chair of the PyTorch Ecosystem Working Group, highlights five talks…
Anthropic floats bill to let U.S. block risky AI models - Briefs Finance — Anthropic floats bill to let U.S. block risky AI models Briefs Finance
DeepSeek的“崩后上新”：宕机5小时，掏出了更高效的大脑 - 搜狐 — DeepSeek的“崩后上新”：宕机5小时，掏出了更高效的大脑搜狐
🔻最新大模型斩杀线。 #deepseekv4.1flash更新##热点聚焦# - 新浪网 — 🔻最新大模型斩杀线。 #deepseekv4.1flash更新##热点聚焦# 新浪网

AI 信息雷达 2026-10-17

Sat, 17 Oct 2026 00:00:00 +0000

必看

Kling AI Officially Launches Next-Generation Kling 4.0, Raising the Bar for Professional-Grade Video Models - Yahoo Finance — Kling AI Officially Launches Next-Generation Kling 4.0, Raising the Bar for Professional-Grade Video Models Yahoo Finance
Kling AI Launches Kling 4.0 Globally with Professional-Grade Output, Stable Motion and Multimodal Control - Yahoo Finance — Kling AI Launches Kling 4.0 Globally with Professional-Grade Output, Stable Motion and Multimodal Control Yahoo Finance
全球首款！中国团队把游戏世界模型塞进一张5090，还开源了 - 风闻 — 全球首款！中国团队把游戏世界模型塞进一张5090，还开源了风闻
Kling 4.0 Released: Why Has Kling AI's Valuation Surpassed Its Parent Company Kuaishou? - 36Kr — Kling 4.0 Released: Why Has Kling AI's Valuation Surpassed Its Parent Company Kuaishou? 36Kr
Kling AI Launches Kling 4.0 Globally with Professional-Grade Output, Stable Motion and Multimodal Control - The Manila Times — Kling AI Launches Kling 4.0 Globally with Professional-Grade Output, Stable Motion and Multimodal Control The Manila Times
金溢科技携手北航发布路侧视角交通世界模型TWM - 观点网 — 金溢科技携手北航发布路侧视角交通世界模型TWM 观点网
黎阳之光发布首创VideoWorld实时世界模型|短视频|实景|仿真|工程|范式 - 手机新浪网 — 黎阳之光发布首创VideoWorld实时世界模型|短视频|实景|仿真|工程|范式手机新浪网
给AI视频生成器配一个懂事的助手：VideoGen-Agent是怎么把翻车现场变成及格线的 - 科技行者 — 给AI视频生成器配一个懂事的助手：VideoGen-Agent是怎么把翻车现场变成及格线的科技行者
可灵AI全量上线Kling 4.0，单次可生成30秒视频 - 三易生活 — 可灵AI全量上线Kling 4.0，单次可生成30秒视频三易生活
昆仑万维开源视频生成模型，单张消费级显卡可实时生成游戏画面 - 搜狐 — 昆仑万维开源视频生成模型，单张消费级显卡可实时生成游戏画面搜狐

值得看

v0.32.0rc3: [Bugfix] Fix NaN output of the Mamba2 chunk scan kernel on sm90 (#61718) — Signed-off-by: Vadim Gimpelson vadim.gimpelson@gmail.com Co-authored-by: Claude noreply@anthropic.com
Anthropic quietly pitches its own AI safety bill to lawmakers, Politico reports - Reuters — Anthropic quietly pitches its own AI safety bill to lawmakers, Politico reports Reuters
【热点聚焦】国内首个！通用世界模型北京市重点实验室在海淀揭牌 - 搜狐 — 【热点聚焦】国内首个！通用世界模型北京市重点实验室在海淀揭牌搜狐
Replit CEO Amjad Masad says company no longer working with Anthropic - Crypto Briefing — Replit CEO Amjad Masad says company no longer working with Anthropic Crypto Briefing
Claude Could Lead 80-90% of Anthropic’s Model Work in 2027 - FourWeekMBA — Claude Could Lead 80-90% of Anthropic’s Model Work in 2027 FourWeekMBA
Claude Mythos 5 对比 DeepSeek V4 Pro：15 倍价格差距，而只有其中一个在自身擅长之处才称得上便宜 - OrcaRouter — Claude Mythos 5 对比 DeepSeek V4 Pro：15 倍价格差距，而只有其中一个在自身擅长之处才称得上便宜 OrcaRouter
xAI’s Grok Voice 3.0 Brings AI Phone Agents To $0.08 A Minute - Minonline — xAI’s Grok Voice 3.0 Brings AI Phone Agents To $0.08 A Minute Minonline
征服“最魔鬼环岛”，奇瑞JAECOO与Momenta智驾无接管穿越凯旋门|巴黎车展|世界模型|sienna|车辆|通行 - finance.sina.com.cn — 征服“最魔鬼环岛”，奇瑞JAECOO与Momenta智驾无接管穿越凯旋门|巴黎车展|世界模型|sienna|车辆|通行 finance.sina.com.cn
You could have invented transformers from memory —
How Itaú is building generative AI for its entire customer base — With ia.i, a conversational interface, and the IARA governance platform, Itaú is applying its AI First strategy to customers and employees — in one of Brazil’s…

AI 信息雷达 2026-10-16

Fri, 16 Oct 2026 00:00:00 +0000

必看

基于动态协同扩散的文本-图像-音频到视频生成框架（DyCoDiff） - 生物通 — 基于动态协同扩散的文本-图像-音频到视频生成框架（DyCoDiff）生物通
快手可灵4.0正式全量上线 - 富途牛牛 — 快手可灵4.0正式全量上线富途牛牛
快手可灵AI模型Kling 4.0全量上线，单次生成视频最长可达30秒 - 搜狐 — 快手可灵AI模型Kling 4.0全量上线，单次生成视频最长可达30秒搜狐
可灵4.0深夜上线，支持30秒一镜到底，熟悉的可灵回来了 - 凤凰网科技 — 可灵4.0深夜上线，支持30秒一镜到底，熟悉的可灵回来了凤凰网科技
可灵AI发布Kling 4.0：30秒原生长视频、15项全能参考、电影级运镜与HDR画质全面升级 - SmartHey — 可灵AI发布Kling 4.0：30秒原生长视频、15项全能参考、电影级运镜与HDR画质全面升级 SmartHey
Kling 4.0正式上线：10-bit HDR、30秒原生直出刷新视频模型新基准 - 爱范儿 — Kling 4.0正式上线：10-bit HDR、30秒原生直出刷新视频模型新基准爱范儿
全球首款！中国团队把游戏世界模型塞进一张5090，还开源了 - 智源社区 — 全球首款！中国团队把游戏世界模型塞进一张5090，还开源了智源社区
nsfw image to video generation new hentai releases for adults 18+ with privacy controls - Universidad de Córdoba (UCO) — nsfw image to video generation new hentai releases for adults 18+ with privacy controls Universidad de Córdoba (UCO)
Meta Ships Muse Wonder, Jev Raises $870M 24 Days After Launch, and Anthropic Cuts Haiku 90%: The Top Takeaways From This Week’s 20VC x SaaStr - saastr.com — Meta Ships Muse Wonder, Jev Raises $870M 24 Days After Launch, and Anthropic Cuts Haiku 90%: The Top Takeaways From This Week’s 20VC x SaaStr saastr.com
FIS and Anthropic launch AI agent to speed up U... - Pluang — FIS and Anthropic launch AI agent to speed up U... Pluang

值得看

停工近5小时，“DeepSeek崩了”炸上热搜，性能之外国产大模型还需补上“稳定课” - 搜狐 — 停工近5小时，“DeepSeek崩了”炸上热搜，性能之外国产大模型还需补上“稳定课” 搜狐
v0.5.22 — Release v0.5.22
Anthropic grants outside evaluator METR employee-level access in AI safety push - WSAU — Anthropic grants outside evaluator METR employee-level access in AI safety push WSAU
Anthropic gives six METR evaluators employee-like access — here are the limits - OpenTools — Anthropic gives six METR evaluators employee-like access — here are the limits OpenTools
create your girlfriend body text to video porn for adults 18+ with privacy controls - Universidad de Córdoba (UCO) — create your girlfriend body text to video porn for adults 18+ with privacy controls Universidad de Córdoba (UCO)
ai porn interactive video AI Porn Documentary for Consenting Adults With Age Verification - Universidad de Córdoba (UCO) — ai porn interactive video AI Porn Documentary for Consenting Adults With Age Verification Universidad de Córdoba (UCO)
ai porn video generation sites Using AI To Create Porn for Consenting Adults With Age Verification - Universidad de Córdoba (UCO) — ai porn video generation sites Using AI To Create Porn for Consenting Adults With Age Verification Universidad de Córdoba (UCO)
Granite Switch 4.1: Chat, Decisions, and Speech from One Endpoint —
free text to video ai porn Is There AI That Can Make Porn for Consenting Adults With Age Verification - Universidad de Córdoba (UCO) — free text to video ai porn Is There AI That Can Make Porn for Consenting Adults With Age Verification Universidad de Córdoba (UCO)
AI公司「月之暗面」IPO 不利消息頻傳 - 大紀元 — AI公司「月之暗面」IPO 不利消息頻傳大紀元

AI 信息雷达 2026-10-15

Thu, 15 Oct 2026 00:00:00 +0000

必看

全球首款！中国团队把游戏世界模型塞进一张5090，还开源了 - 凤凰网科技 — 全球首款！中国团队把游戏世界模型塞进一张5090，还开源了凤凰网科技
Introducing Grok Voice 3.0 | SpaceXSI - xAI — Introducing Grok Voice 3.0 | SpaceXSI xAI
v1.4.0rc0 — Highlights Known Issues Multi-GPU (TP2) streaming or RPC executor workloads may hang during worker initialization when MPI worker pools are reused across LLM…
Amp puts AI agents to work beyond code with OpenAI models — See how Amp pushes the frontier of what agents can do with GPT-6 Astra, Ultrafast, GPT-Live-1, and Sign in with ChatGPT.
Choosing AI Image and Video Generation Tools for Adult - Universidad de Córdoba (UCO) — Choosing AI Image and Video Generation Tools for Adult Universidad de Córdoba (UCO)
京东工业发布工业世界模型，与AMD携手推动模型与硬件深度融合 - 搜狐 — 京东工业发布工业世界模型，与AMD携手推动模型与硬件深度融合搜狐
阶跃终端首款「Agent手机」STEPX Neo正式发布，创客先行 — 这妥妥是个物理版Agent啊！
Seedance 3.0 Brings Feature-Length Control to AI Video Generation - Startup Fortune — Seedance 3.0 Brings Feature-Length Control to AI Video Generation Startup Fortune
How Third Moment Research runs 10,000 agents for 36+ hours — With GPT-6 Astra and GPT-6.1 Sol, Codex, and Daybreak for cybersecurity, this AI lab builds agents that focus for days and improve with use.
STEPX首款大模型原生智能体手机STEPX Neo正式发布 — 2026年10月13日，大模型原生AI终端品牌STEPX（阶跃终端）旗下首款大模型原生智能体手机STEPX Neo正式发布

值得看

v0.32.0rc1: [Dependency] Update Humming and remove NVFP4 scale workaround (#61556) — Signed-off-by: jinzhen.ljz jinzhen.ljz@antgroup.com Co-authored-by: Codex codex@openai.com Co-authored-by: Misha Goin mgoin64@gmail.com Co-authored-by: Kevin…
苹果iPhone Duo国行备案新增大模型编号：Apple智能、通义万象、通义千问 - 新浪财经 — 苹果iPhone Duo国行备案新增大模型编号：Apple智能、通义万象、通义千问新浪财经
影响市场重大事件：2026物理AI与世界模型大会将于10月19日至22日在南京举行；美政府开始审理SpaceX星舰扩张提案；交通运输部印发《水运领域人工智能应用场景指引》 - https://www.mrjjxw.com/ — 影响市场重大事件：2026物理AI与世界模型大会将于10月19日至22日在南京举行；美政府开始审理SpaceX星舰扩张提案；交通运输部印发《水运领域人工智能应用场景指引》 https://www.mrjjxw.com/
Crossing the Threshold: How OpenAI and Anthropic Approach Autonomy and Safety in Financial Agents — At AI Engineer New York, frontier labs and financial firms explored how to give agents useful autonomy while keeping consequential decisions under control.
Lab-in-the-Loop at Compute Speed — Ian Quigley, Leash Bio — How Leash Bio tamed DELs, the assay that traumatizes everyone, and used the data to catch cheating models…
Take a Look Under the Hood? The Nuts and Bolts of How Spacial’s Construction World Model Will Change Building - Mann Publications — Take a Look Under the Hood? The Nuts and Bolts of How Spacial’s Construction World Model Will Change Building Mann Publications
【网经社AI科普】阶跃星辰（跃问）：专注多模态研发推理与视觉感知突出的大模型 - 搜狐 — 【网经社AI科普】阶跃星辰（跃问）：专注多模态研发推理与视觉感知突出的大模型搜狐
A Practical Guide to Seedance 2.5 API Access for Startups and Developer Teams - nerdbot — A Practical Guide to Seedance 2.5 API Access for Startups and Developer Teams nerdbot
Shaquille O’Neal Shows Off Shirtless Workout, Channels Arnold Schwarzenegger in Viral Video - Generation Iron — Shaquille O’Neal Shows Off Shirtless Workout, Channels Arnold Schwarzenegger in Viral Video Generation Iron
企业真正需要的，不是更聪明的 Agent ？ — 点击查看原文>

AI 信息雷达 2026-10-14

Wed, 14 Oct 2026 00:00:00 +0000

必看

京东工业发布工业世界模型助力工业现场具备未来感知能力 - 搜狐 — 京东工业发布工业世界模型助力工业现场具备未来感知能力搜狐
v2.8.3.post2 — Maintenance release of FlashAttention 2 on top of v2.8.3.post1. No kernel feature changes. Backport #2622 : torch.load(..., weights_only=True) in…
T早报｜字节跳动单独设立大模型安全部门；阶跃星辰推出智能体手机；美国拟投票封杀认证中国电子设备测试机构 - 财新 — T早报｜字节跳动单独设立大模型安全部门；阶跃星辰推出智能体手机；美国拟投票封杀认证中国电子设备测试机构财新
微信上线绿泡泡短剧|AI视频生成|腾讯|红果|广告|免费 - finance.sina.com.cn — 微信上线绿泡泡短剧|AI视频生成|腾讯|红果|广告|免费 finance.sina.com.cn
康冠科技：近期已推出搭载阶跃星辰大模型的闺密机、首款mini AI PC等产品 - 东方财富 — 康冠科技：近期已推出搭载阶跃星辰大模型的闺密机、首款mini AI PC等产品东方财富
阶跃星辰智能体手机推出和其他智能体手机有何不同？ - 财新 — 阶跃星辰智能体手机推出和其他智能体手机有何不同？财新
v13.4.4: Prepare the 13.4.4 release (#3096) — Prepare the 13.4.4 release More complete removal of Python 3.15 pre-release status
Anthropic 发布蒸馏防御研究，点名智谱和 Kimi - Pasquale Pillitteri — Anthropic 发布蒸馏防御研究，点名智谱和 Kimi Pasquale Pillitteri
Waters Launches AI Link, Advancing AI-Enabled LC-MS Workflows from Research to Release - The Manila Times — Waters Launches AI Link, Advancing AI-Enabled LC-MS Workflows from Research to Release The Manila Times
Extending Inductor for the IBM Spyre Accelerator — TL;DR The PyTorch-based software stack for IBM’s Spyre accelerator recently moved its torch.compile backend from sendnn, a separate compilation stack built…

值得看

【首发】高维医药完成Pre-A轮融资，加速构建AI靶点发现世界模型 - 动脉网 — 【首发】高维医药完成Pre-A轮融资，加速构建AI靶点发现世界模型动脉网
llm-gemini 0.35 — Release: llm-gemini 0.35 I added support to llm-gemini for the new Gemini 4 Argon model, made available today via the Gemini API and for customers of their…
Dyna-2: General Purpose Physical Agents with Hour-Long Autonomy — Jason Ma, Dyna Robotics — Our first robotics pod with Dyna’s Chief Scientist on their big bet on World Action Models & the Next Frontier of Physical AI
How to Rewrite Apache Spark UDFs for GPU Acceleration with AI Agents — Enterprise applications have become increasingly demanding. Agentic AI has compounded the need for vast amounts of enterprise data with frequent, complex...
Anthropic Demonstrates AI Worm Components That Could Spread Without Data Centers, But Practical Threat Remains Years Away - news.lavx.hu — Anthropic Demonstrates AI Worm Components That Could Spread Without Data Centers, But Practical Threat Remains Years Away news.lavx.hu
Anthropic Says New Defenses Cut Data Harvested Through AI Distillation Attacks By 90% To 99% - Pulse 2.0 — Anthropic Says New Defenses Cut Data Harvested Through AI Distillation Attacks By 90% To 99% Pulse 2.0
刚亮相就传被英伟达收购：Gemini老兵团队凭什么还敢押注Coding Agent？ — 点击查看原文>
Claude 开始自己写程序管理 Agent 了：Anthropic 披露动态工作流架构 — 点击查看原文>
Run fast, private agents on-device with LFM2.5-24B-A2B —
AI大模型聚合站供应商怎么挑？看完这篇少踩坑多省钱 - CSDN — AI大模型聚合站供应商怎么挑？看完这篇少踩坑多省钱 CSDN

AI 信息雷达 2026-10-13

Tue, 13 Oct 2026 00:00:00 +0000

必看

跨维智能发布全球首个3D原生世界模型UNITAS，统一建模动作与世界变化 - SmartHey — 跨维智能发布全球首个3D原生世界模型UNITAS，统一建模动作与世界变化 SmartHey
阶跃星辰选在开源前夜涨价 - 中华网财经 — 阶跃星辰选在开源前夜涨价中华网财经
智谱AI发布法律垂直大模型GLM-Legal 基于GLM-5.3-Flash架构 - 观点网 — 智谱AI发布法律垂直大模型GLM-Legal 基于GLM-5.3-Flash架构观点网
Intel CEO肯定中国开源权重模型：DeepSeek、Kimi、GLM“已足够好” - 디지털투데이 — Intel CEO肯定中国开源权重模型：DeepSeek、Kimi、GLM“已足够好” 디지털투데이
Meta claims reaching AI 'frontier' with release of 2 new models - Daily Sabah — Meta claims reaching AI 'frontier' with release of 2 new models Daily Sabah
Meta releases Muse Wonder AI models to challenge Anthropic and OpenAI - Reuters — Meta releases Muse Wonder AI models to challenge Anthropic and OpenAI Reuters
Launch HN: Agent37 (YC F26) – Low-cost agent sandboxes for running agents — Hey HN, I’m Vishnu from Agent37 ( https://agent37.com ). We build sandboxes for running AI agents at under 10% of the cost of most sandbox providers…
Reveal Technology Launches Farsight 2.0 With a Global World Model for the Tactical Edge - Yahoo Finance — Reveal Technology Launches Farsight 2.0 With a Global World Model for the Tactical Edge Yahoo Finance
黄仁勋的下一枚棋子：英伟达收编开源模型，把算力需求焊死在GPU上|DeepSeek|Reflection AI|芯片|DeepMind|Kimi - finance.sina.com.cn — 黄仁勋的下一枚棋子：英伟达收编开源模型，把算力需求焊死在GPU上|DeepSeek|Reflection AI|芯片|DeepMind|Kimi finance.sina.com.cn
知天下3DGS社区开源世界模型基模Explorer 1.3B - 砍柴网 — 知天下3DGS社区开源世界模型基模Explorer 1.3B 砍柴网

值得看

MiniMax入驻天猫开卖Token套餐，多家大模型厂商跟进 - 凤凰网科技 — MiniMax入驻天猫开卖Token套餐，多家大模型厂商跟进凤凰网科技
MiniMax入驻天猫开设首家品牌官方旗舰店上架M Plan三档订阅套餐 - 观点网 — MiniMax入驻天猫开设首家品牌官方旗舰店上架M Plan三档订阅套餐观点网
全球首个机器人任务原生3D世界模型来了 - DoNews — 全球首个机器人任务原生3D世界模型来了 DoNews
【视频】不挑路况/德系底盘/世界模型/极致安全，ID. ERA 8X 20.68万元起 - 汽车之家 — 【视频】不挑路况/德系底盘/世界模型/极致安全，ID. ERA 8X 20.68万元起汽车之家
2026物理AI与世界模型大会将在南京举行，观众可近距离感受人形机器人、自动驾驶 - 新浪财经 — 2026物理AI与世界模型大会将在南京举行，观众可近距离感受人形机器人、自动驾驶新浪财经
黄仁勋再度点赞中国大模型：DeepSeek、Kimi、GLM表现足够优秀 - 极客网 — 黄仁勋再度点赞中国大模型：DeepSeek、Kimi、GLM表现足够优秀极客网
Kling AI: Next-Generation AI Tool at Artlist - Character Media — Kling AI: Next-Generation AI Tool at Artlist Character Media
Intel D270 GPU AI Accelerator–Up to 480 GB LPDDR5X —
OCP Summit 2026: Powering AI with Open Innovations in Scale-Up, Scale-Out, and Scale-Across Networks — At the 2026 Open Compute Project Global Summit, Meta is sharing the next generation of open networking infrastructure, including: Our first liquid-cooled…
OpenAI Negotiates to Restart Sora Video Model Collaboration - https://www.ababnews.com/ — OpenAI Negotiates to Restart Sora Video Model Collaboration https://www.ababnews.com/

AI 信息雷达 2026-10-12

Mon, 12 Oct 2026 00:00:00 +0000

必看

H-JEPA！LeCun世界模型创业交卷，代码权重全部开源 - 搜狐网 — 搜狐：LeCun 世界模型创业公司开源 H-JEPA，代码与权重全部公开
量化后画面闪烁？时序稳定的视频世界模型2-bit KV Cache压缩方案诞生 - 搜狐 — 搜狐：时序稳定视频世界模型 2-bit KV Cache 压缩方案诞生
中国科学院工业人工智能研究所PAIWorld将在2026物理AI与世界模型大会发布 - 新华网江苏频道 — 新华网：中科院工业人工智能研究所 PAIWorld 将在世界模型大会发布

值得看

综合Ⅱ行业点评报告：港股消费&应用：阜博集团收购Dreamesh完善AI视频生成版图，东方甄选重视自营产品建设 - 证券之星 — 证券之星：阜博集团收购 Dreamesh 完善 AI 视频生成版图
综合行业周报：海外科技：AMD布局世界模型，模型能力升级与成本优化同步推进_行业研究_研报 - 证券之星 — 证券之星研报：AMD 布局世界模型，模型能力升级与成本优化并行
Amazon is reportedly circling Decart after Nvidia and Anthropic both walked away - Startup Fortune — Startup Fortune：Amazon 据报正接触 Decart，Nvidia 与 Anthropic 已退出
Amazon in talks to acquire AI startup Decart for about $7 billion - Crypto Briefing — Crypto Briefing：Amazon 洽谈以约 70 亿美元收购 Decart
New Sakana AI paper: Recursive Self-Improvement through Multi-Agent Self-Supervision (MASS). Blog: pub.sakana.ai/mass/ P — Sakana AI 发布 MASS 论文：通过多智能体自监督实现递归自我改进。
“L1相对成熟、L2已见曙光”：通用世界模型的发展卡在哪里？|清华大学|北京市|数据|实验室|芯片 - 新浪财经 — 新浪财经：清华等讨论通用世界模型发展瓶颈，L1 成熟、L2 现曙光
Anthropic Is Cutting AI Agents Off From the Live Internet - WeRSM — Anthropic 切断其 AI Agent 的实时互联网访问。
《华盛顿邮报》：美国争夺AI主导权，却离不开中国模型和制造业供应链 - 加美财经 — 华盛顿邮报：美国争夺 AI 主导权但依赖中国模型与制造业供应链。
国内首个世界模型方向的省部级重点实验室在京揭牌，极佳视界为依托单位 - 新浪财经 — 新浪财经：国内首个世界模型省部级重点实验室揭牌，极佳视界为依托单位
Odyssey-3 is a new generative world model that you can try for free - The Decoder — The Decoder：Odyssey-3 新生成式世界模型开放免费试用

AI 信息雷达 2026-10-11

Sun, 11 Oct 2026 00:00:00 +0000

必看

Running large Mixture-of-Experts models across pods creates two major challenges: expert traffic and KV-cache transfer. — PyTorchCon：MoRI + vLLM 解决跨 pod MoE 模型的专家流量与 KV-cache 传输。
“视频大模型第一股”角逐：可灵AI拟赴港IPO，最快明年启动 - 新浪网 — 快手可灵 AI 拟赴港 IPO，最快 2027 年启动
H-JEPA！LeCun世界模型创业交卷，代码权重全部开源 - 搜狐 — LeCun 团队 H-JEPA 世界模型开源，代码权重全部公开
Shengshu Technology Launches Vidu Q4 Preview Version with Support for 4K Output - AIBase — 生数科技发布 Vidu Q4 Preview，支持 4K 输出
Odyssey发布Odyssey-3系列基础世界模型，Pro版刷新Physics-IQ Verified视频预测纪录 - SmartHey — Odyssey-3 Pro 刷新 Physics-IQ Verified 视频预测纪录
DeepSeek联手华为开源全家桶，英伟达CUDA的墙被撬开了 - 手机网易网 — DeepSeek 联手华为开源全家桶，挑战英伟达 CUDA 生态。
Pumpire: Unified Benchmark for Metric Distance Estimation — Pumpire：面向 3D 基础模型的统一基准，直接评估点-点距离估计能力，含 100 个真实场景。
One Block, Multiple Depths: Recurrent Vision Transformers with Depth-Programmed Experts — reViT：单个 Transformer block 循环复用 + 深度编程专家混合，以低 FLOPs 匹配全深度 ViT 精度。
SpaceFlow: Locally Controllable 3D Generation — SpaceFlow：基于 flow 的免训练局部可控 3D 生成管线，通过几何 primitive 实现区域级控制。
WorldGuide: Goal-Directed Video World Model for Procedural Task Execution — WorldGuide：面向程序化任务执行的目标导向视频世界模型，闭环生成与执行。

值得看

We've been monitoring a decline in HuggingFace downloads across the top open LLMs we track for interconnects. It looks l — Nathan Lambert 称 HuggingFace 热门模型下载量下降约 30%。
Google's Gemini 4 "Carbon" model reportedly feels like Anthropic's Opus 5.5 coding performance - The Decoder — Google Gemini 4 "Carbon" 模型据称编码能力接近 Anthropic Opus 5.5。
物理世界模型的Scaling Law：真正稀缺的不是机器人而是可规模化的物理数据 - 驱动之家 — 物理世界模型 Scaling Law：稀缺的是可规模化物理数据
Anthropic Cuts Live Internet Access for Internal AI Tests After Claude Exploits Injection Flaws - The Hacker News — Anthropic 在 Claude 利用注入漏洞后切断内部 AI 测试的实时互联网访问。
Anthropic Launches Free OSS Scanner to Find AI-Detected Software Vulnerabilities - Konsulteer — Anthropic 发布免费开源软件漏洞扫描工具。
Odyssey-3 基础世界模型登场 - DoNews — Odyssey-3 基础世界模型发布
世界模型技术路径观察：四类方向与代表企业梳理 - CSDN — 世界模型四类技术路径与代表企业梳理
让自然语言成为世界模型的表征，英伟达Physis-Lang这样增强视频物理真实性 - 搜狐网 — 英伟达 Physis-Lang：用自然语言表征增强视频物理真实性
MiniMax跌掉80%、DeepSeek拟融资1000亿：大模型IPO潮谁来接盘？-36氪 - 36kr — MiniMax 跌 80%、DeepSeek 拟融资 1000 亿，大模型 IPO 潮引关注。
Anthropic can't reliably control its AI agents. It's cutting off its internal evals from the live internet instead - TechCrunch — Anthropic 因无法可靠控制 AI agent，切断内部评估与公网的连接。

AI 信息雷达 2026-10-10

Sat, 10 Oct 2026 00:00:00 +0000

必看

生数科技发布视频生成模型Vidu Q4 Preview预览版本 - 凤凰网科技 — 生数科技发布视频生成模型 Vidu Q4 Preview 预览版。
@ 一下就能派活？谷歌推出办公 Agent，拥有独立账号、能够创建子 Agent，还能调用 Claude — 点击查看原文>
Impactful scheduling for GPU clusters — Hugging Face 博客发布 GPU 集群调度优化文章。
Anthropic Launches Claude Haiku 5.5 as OpenAI Rolls Out GPT-6 and Interactive ChatGPT Features - Tekedia — Anthropic Launches Claude Haiku 5.5 as OpenAI Rolls Out GPT-6 and Interactive ChatGPT Features Tekedia
Fluid-Gen-Zero: Grounding Pretrained Video Generators in Physics without Training — Fluid-Gen-Zero：无需训练即可将物理仿真注入预训练视频生成器。
iCATS: Fast Video Generation via Interaction-Aware Sparse Attention and Timestep-Adaptive Sparsity — iCATS：通过交互感知稀疏注意力实现快速视频生成。
Conditional Residual Prediction: Improving Autoregressive Video Diffusion without a Bidirectional Teacher — 提出 Conditional Residual Prediction，仅用图像模型初始化训练因果视频扩散模型，无需双向教师蒸馏。
Parametric Trajectory Distillation for Few-Step Video Generation — 提出 Parametric Trajectory Distillation (PTD)，用多项式参数化教师轨迹段以改善少步视频生成蒸馏。
Memory Forcing: Attendable Mid-Horizon History for Streaming Video Generation — 提出 Memory Forcing，通过 Archive & Working Banks 分区 KV 缓存解决流式视频生成的 mid-horizon 遗忘问题。
Phase-aware video generation for physics-grounded dynamics and interactions — 提出 PAVG，双分支 Phase-Aware Video Generator 分别建模固体/气体动力学并通过交叉注意力耦合。

值得看

Anthropic AI model submits false tip to police website - Cootamundra Herald — Anthropic AI model submits false tip to police website Cootamundra Herald
Anthropic AI model submits false tip to police website - The Canberra Times — Anthropic AI model submits false tip to police website The Canberra Times
Anthropic under fire as AI model submits false homicide tip to Philadelphia police | Hindustan Times - Hindustan Times — Anthropic under fire as AI model submits false homicide tip to Philadelphia police | Hindustan Times Hindustan Times
I’m more bought into AI becoming far more efficient, which is a massive lever on diffusion, than I am on AI models makin — I’m more bought into AI becoming far more efficient, which is a massive lever on diffusion, than I am on AI models making massive breakthroughs that result in…
Anthropic AI model submits false homicide tip to police website - CTV News — Anthropic AI model submits false homicide tip to police website CTV News
Google is about to roll out a new AI model. Employees say they're testing another that's way better. - Business Insider — Google 即将推出新 AI 模型，员工称正在测试另一款更强的模型。
Connect with Madrid’s PyTorch community on October 14, 6:00–8:30 pm, at a free meetup to share real-world experiences an — Connect with Madrid’s PyTorch community on October 14, 6:00–8:30 pm, at a free meetup to share real-world experiences and discuss technical questions…
AMD收购World Labs 补齐AI生态布局 - 搜狐网 — AMD 收购 World Labs 补齐 AI 生态布局。
High-Velocity GPU Kernel Authoring with CUTLASS Python At PyTorch Conference North America 2026, Guray Ozen from NVIDIA — High-Velocity GPU Kernel Authoring with CUTLASS Python At PyTorch Conference North America 2026, Guray Ozen from NVIDIA will present recent improvements to…
从 Demo 到生产：AI Agent 缺的到底是什么？ — 点击查看原文>

AI 信息雷达 2026-10-09

Fri, 09 Oct 2026 00:00:00 +0000

必看

不等Gemini 4了！谷歌发布办公Agent，支持调用Claude — 谷歌发布办公 Agent，支持调用 Claude 等多模型。
NVIDIA recsys-examples now provides an end-to-end HSTU inference workflow with Dynamo-Triton. PyTorch AOTI compiles HSTU — NVIDIA recsys-examples 提供 HSTU 端到端推理，Dynamo-Triton + AOTI 延迟降 5.93x。
Optimizing kernels can be time-consuming and require deep expert knowledge. KernelAgent is a multi-agent harness that ca — Meta 提出 KernelAgent：多 Agent 自动编写和优化 GPU kernel。
Session-Aware Agentic Inference with NVIDIA Dynamo — PyTorch Blog：NVIDIA Dynamo 实现 Session-Aware Agentic 推理。
Step 5 Preview, a 1M-context MoE from StepFun, shows up on OpenRouter — 阶跃星辰 Step 5 Preview（1M 上下文 MoE）现身 OpenRouter。
Odyssey Launches Odyssey-3, Its Most Powerful Foundation World Model Built to Power AI Across the Physical World - Business Wire — Odyssey发布Odyssey-3，其最强基础世界模型，支持物理世界AI。
Building Spyre as a Native PyTorch Device — PyTorch Blog：将 Spyre 构建为原生 PyTorch 设备。
零跑汽车发布不依赖BEV的世界模型智驾系统，称迈入行业第一梯队 - 虎嗅网 — 零跑汽车发布不依赖BEV的世界模型智驾系统，称迈入行业第一梯队。
视频模型迎来“DeepSeek时刻”？生数科技发布 Vidu Q4 Preview 高表现力旗舰模型 - 凤凰网科技 — 生数科技发布 Vidu Q4 Preview 高表现力旗舰视频模型。
Vidu Q4 Preview发布：推动旗舰视频模型大规模应用落地 - 搜狐网 — Vidu发布Q4 Preview，推动旗舰视频模型大规模应用落地。

值得看

视频DeepSeek时刻？Vidu Q4 Preview背后的效率账与世界模型底气 - 搜狐网 — 生数科技 Vidu Q4 Preview 引发视频 DeepSeek 时刻讨论。
代码造世界，扩散绘现实：AgentGarten让智能体在实时试炼场中边玩边进化 — AgentGarten 提出让智能体在实时试炼场中边玩边进化的框架。
Anthropic Adds Dashboards and Video Generation to Claude, Refreshes Small Model Lineup - BigGo Finance — Anthropic 为 Claude 新增 Dashboards 和视频生成功能，更新小模型阵容。
Read our latest blog from Nvidia to learn how NVIDIA Dynamo integrates its router, inference engine, and KV-cache manage — NVIDIA Dynamo 集成路由、推理引擎和 KV-cache 管理，服务 Agent 负载。
Odyssey-3 drives a car in India and runs humanoids from one world model - The Next Web — Odyssey-3在印度驾驶汽车并运行人形机器人，展示世界模型能力。
OpenAI刚把多Agent做成产品，o1奠基人却说：1万个Agent解出世界级难题，多Agent贡献不到10% — OpenAI o1 奠基人质疑多 Agent 效果：1 万 Agent 解难题，多 Agent 贡献不到 10%。
Mistral opens trillion-parameter Large 4 preview, weights due in October - BetaNews — Mistral 开放万亿参数 Large 4 预览，权重预计 10 月发布。
Former ByteDance intern Tian Keyu secures $30 million in funding; world model startup reaches $200 million valuation - BigGo Finance — 前字节实习生田柯宇创办世界模型公司获3000万美元融资，估值2亿美元。
How can a single framework support an increasingly diverse range of AI hardware? In a new PyTorch Foundation blog, contr — IBM 分享 Spyre 集成到 PyTorch 的抽象层方案，支持多元 AI 硬件。
北大博士世界模型初创公司获3000万美元融资，视频生成成本压缩至十分之一|赛道|仿真|训练|视觉|推理 - 手机新浪网 — 北大博士世界模型初创公司获 3000 万美元融资，视频生成成本压缩至十分之一。

AI 信息雷达 2026-10-08

Thu, 08 Oct 2026 00:00:00 +0000

必看

Anthropic Releases Claude Haiku 5.5: A Small Model With 1M Context Priced at $0.10 per Million Input Tokens - MarkTechPost — Anthropic 发布 Claude Haiku 5.5，支持 1M 上下文，输入价格 $0.10/M tokens。
Shengshu Technology Launches Vidu Q4 Preview, A Next-Generation AI Video Model Built for Lifelike Performances - GlobeNewswire — 生数科技发布 Vidu Q4 Preview 新一代 AI 视频模型。
Artemis: Geometry-Grounded Multi-Agent Driving World Models with Shared 3D State and Progressive Memory Update — arXiv:2610.07031v1 Announce Type: new Abstract: Recent video world models have witnessed the paradigm shift from single-agent to multi-agent involvements,…
GeoWM: Efficient Direct World Modeling in Explicit Geometry — arXiv:2610.07381v1 Announce Type: new Abstract: Modeling 3D scene geometry and its evolution over time is essential for autonomous driving and robotics…
CCDF: A Benchmark Dataset for Deepfake Detection in Real-World Surveillance Footage — arXiv:2610.07939v1 Announce Type: new Abstract: Due to rapid advances in Generative AI, commercial video generation tools can be used to produce fabricated…
Backend-Agnostic Sparse Attention for Fast High-Resolution Visual Generation — arXiv:2610.08772v2 Announce Type: new Abstract: Diffusion Transformers (DiTs) have achieved strong performance in image and video generation, but the quadratic…
Rethinking Visual Provenance: Detection and Watermarking Across Direct Visual Generation and LLM-Driven Code Rendering — arXiv:2610.08137v1 Announce Type: cross Abstract: AI systems create images and videos with image/video generation models or by writing code and graphics…
Learning Visual Feature-Based World Models via Residual Latent Action — arXiv:2605.07079v2 Announce Type: replace Abstract: World models predict future transitions from observations and actions…
ACG-WAM: World-Action Modeling via Action-Conditioned Geometric Latent Prediction — arXiv:2610.06965v1 Announce Type: new Abstract: World action models jointly learn visual predictionand robot actions, providing a way to use observations…
对标中国AI模型：英伟达投资的Reflection AI发布旗下首款开放权重模型Beam|Kimi|Qwen3.8|智谱|DeepSeek|Qwen 3 - 新浪财经 — 英伟达投资的 Reflection AI 发布首款开放权重模型 Beam，对标中国 AI 模型。

值得看

港股大模型“双雄”走弱 MINIMAX-W一度跌超10% - 东方财富 — 港股大模型双雄走弱，MINIMAX-W 一度跌超 10%。
字节跳动前实习生田柯宇获近3000万美元融资，创办世界模型公司叫板李飞飞 - 凤凰网 — 字节跳动前实习生田柯宇获近3000万美元融资，创办世界模型公司叫板李飞飞凤凰网
DeepSeek新一轮融资或达千亿元；快手可灵AI计划最早明年赴港上市｜未来商业早参 - https://www.mrjjxw.com/ — DeepSeek 新一轮融资或达千亿元；可灵 AI 计划明年赴港上市。
Mistral Just Revealed a 1-Trillion-Parameter AI Model. It Has a Very Funny Nickname - Inc.com — Mistral 发布 1 万亿参数模型，绰号 Le Chonk。
Docker Agent — Docker Agent 登上 Hacker News 热榜（100+ 分）。
Multimodal open d1 decision models for the edge — Hugging Face 博客介绍面向边缘设备的开源多模态决策模型。
消息称DeepSeek接近完成至少800亿元融资，腾讯、宁德时代重金参与 - 新浪网 — 消息称 DeepSeek 接近完成至少 800 亿元融资，腾讯、宁德时代参与。
Can a Cloud-Native Harness Make Agents Reliable Beyond the Desktop? — Latent Space 讨论云原生 harness 能否让 Agent 在桌面外可靠运行。
One Model Family, Two Gold-Level Results: Fine-Tuning Nemotron for IOI and IMO — NVIDIA Nemotron 微调后在 IOI 和 IMO 获得金牌级成绩。
《怪物史莱克》编剧也来了！这家AI影视公司，视频模型全球第二！ — 某 AI 影视公司视频模型全球第二，聘请《怪物史莱克》编剧。