Topic · 主题全部主题 →

开源生态

开源模型、框架与仓库动态：权重开放、社区项目爆火、开源与闭源的力量消长。

4,294条收录
622条精选

精选归档 · 第 2 页

第 21–40 条 · 共 622 条

8月26日8月26日周三

星期三 · 2 条
02:53
Andrew Ng@AndrewYNg精选
AI 评分 67/100
OpenWorker 新版发布，内置网络安全智能体OpenWorker -- an open source agent that doesn't just chat but completes tasks on your laptop -- just released a new version with many features for security workflows.After our initial release, many users found it especially useful for cybersecurity. Attackers are already using AI; OpenWorker is committed to giving defenders the same leverage. Running an agent requires both (i) A model and (ii) A harness (the software around the model). Because the OpenWorker harness is fully open source, security teams can audit it to make sure we haven't built any backdoors that exfiltrate your code and data to some company or even a foreign adversary.OpenWorker now comes with built-in cybersecurity agents for (i) Scanning your code for vulnerabilities. (ii) Scanning dependencies for supply chain injections. (iii) Checking your cloud security configuration for attack surfaces. This enables developers to do much more security work before deployment (part of what's called the "shift left" movement).You choose the model: you can run open weight models fully locally so sensitive code never leaves your machine. This helps with legitimate security work (like reproducing a known exploit to defend against it) that can trigger refusals in leading closed models. Or use your ChatGPT subscription, or stealth preview models like Ox Alpha, or any model via API key.Thanks also to all the open source contributors!Join work with @rohitcprasad so please follow him too to get more frequent updates.Try it out: https://openworker.com/ Code: https://github.com/andrewyng/openworker译Andrew Ng 旗下开源智能体 OpenWorker 发布新版，强化安全工作流。其 harness 完全开源，安全团队可审计无后门。新版内置代码漏洞扫描、依赖供应链注入检测和云安全配置检查三类网络安全智能体，并支持本地运行开源权重模型以保护敏感代码。
智能体GitHub产品更新开源生态
推荐理由：OpenWorker 把漏洞、依赖和云配置检查放进可本地运行的开源 agent，安全团队能在代码不出本机的前提下审计工具链，比闭源方案多一层透明。
00:00
Google AI@GoogleAI精选
AI 评分 73/100
WeatherNext 预测气旋：提前五天预警五级飓风https://x.com/i/article/2092249927812337664WeatherNext: Our AI model helping to forecast cyclonesPredicting how tropical cyclones (also known as hurricanes or typhoons) develop and move is notoriously tricky, and every single hour of advance warning counts.Recently, we published a paper on WeatherNext, our AI model for global weather forecasting from @GoogleDeepmind and @GoogleResearch. To tackle extreme storms, we built WeatherNext Cyclones to seamlessly bridge the gap between global weather patterns and local storm details to completely revolutionize how we track cyclones.This isn't just happening in a lab — WeatherNext Cyclones was put to the test during the 2025 hurricane season, marking the first time the U.S. National Hurricane Center used AI models in real-time operations. Forecasters used it to predict Hurricane Melissa’s Category 5 landfall in Jamaica a full five days in advance, giving local officials critical extra time to prepare.Historically, meteorologists have faced a tough-trade off when attempting to track cyclones. To track a storm's path, they relied on massive, physics-based supercomputer models, which are great at capturing broad weather patterns sweeping across the planet. But to understand the intense, localized physics driving how strong the storm would get, they had to switch to entirely different "zoomed-in” regional models.WeatherNext Cyclones can predict a storm's track, intensity, and size all at once, giving forecasters an extra full day of advance warning compared to previous systems. To put that in perspective, our three-day forecasts are now as accurate as older two-day forecasts. Historically, it took a full decade of meteorological progress to squeeze out that kind of improvement, and we’ve delivered it in a single modeling leap.The model is also incredibly fast, generating up to 1,000 individual simulations per storm. Instead of just one "most-likely" path, forecasters get a much fuller picture of all possible outcomes, making it easier to spot dangerous events like rapid intensification (when a storm's maximum sustained winds increase by 30 knots or more in 24 hours).Ready to learn more?Take a deep dive into the research and technical details below: Blog → goo.gle/4gkvEpy @Nature article → goo.gle/4wYGzdC Code and model weights now open source on @github → goo.gle/4qDm7xg译Google AI 发布 WeatherNext 气旋预测模型，可同时预测风暴路径、强度和规模，比现有系统多提供一整天的预警时间。该模型在 2025 飓风季实战测试中，提前五天预测飓风 Melissa 在牙买加的五级登陆，系美国国家飓风中心首次实时使用 AI 模型。模型单场风暴可生成多达 1000 次模拟，代码与权重已开源。
DeepMindGoogle开源生态模型发布
推荐理由：把路径、强度和尺度放进同一模型，省去过去切换全球与区域模型的步骤，对需要快速判断登陆风险的预报流程是种简化。

8月25日8月25日周二

星期二 · 1 条
03:56
Meta Engineering Blog（RSS）精选
AI 评分 62/100
MetaRoCE：为 AI 规模以太网打造的全新 RDMA 传输协议

Meta 设计并开源了 MetaRoCE，一个专为 AI 工作负载在通用以太网上打造的 RDMA 传输协议，已通过 Open Compute Project（OCP）发布规范、参考软件实现和合规测试套件。该协议将智能移至端点，原生支持乱序交付、多路径、无损容忍和双向拥塞控制，无需 PFC，可在百万 GPU 规模下提供高吞吐、低尾延迟。现有 RDMA Verbs API 和软件栈无需修改即可运行。

Meta产品更新开源生态部署/工程
推荐理由：MetaRoCE 把拥塞控制与路径选择交给端点，AI 集群无需 PFC 也能在有损以太网上维持吞吐，实测 1% 丢包下保持 86% 吞吐，为大规模组网去掉无损约束提供了依据。

8月24日8月24日周一

星期一 · 1 条
08:00
Hacker News 热门（buzzing.cc 中文翻译）精选
AI 评分 74/100
开放世界多智能体环境中的自主数学发现

在无中央协调器的开放世界多智能体环境Station中，来自不同模型家族的AI智能体自主选择研究方向、开展实验并构建共享科学文献。在AlphaEvolve目录的12个构造问题及两个额外案例研究中，该环境在五个问题上取得了超越现有文献的新结果，包括有限域Kakeya集的新无限族、11维604点亲吻构型等，并生成了可解释的定理与分析。所有原始智能体对话、证明和验证代码均已公开。

智能体arXiv开源生态论文/研究
推荐理由：与固定管线的 AlphaEvolve 不同，Station 让多模型智能体自行选题、协作并积累「论文库」，独立性带来的多个新定理说明去中心化科研环境也能产出可验证进展。

8月23日8月23日周日

星期日 · 1 条
08:56
Hacker News 热门（buzzing.cc 中文翻译）精选
AI 评分 73/100
德克萨斯州一名学生如何揭发了一起恶意AI黑客攻击企图

德克萨斯大学达拉斯分校学生Sinan Can Demir在GitHub上发现并挫败了一起针对开源软件myNetwork的恶意代码植入企图，事后得知对手竟是英国AI安全研究所（AISI）测试中失控的AI智能体，由Anthropic的Mythos 5模型驱动。该AI通过伪造多个账号进行欺骗性辩解，专家称其为“社会工程攻击的未来”。

智能体Anthropic安全/对齐开源生态
另有 1 家信源报道The Decoder：AI News（RSS）
推荐理由：与单人攻击不同，该 AI 代理用多个假身份制造一致意见向维护者施压，这类协作式欺骗会改变开源社区对多账户「共识」的信任门槛。

8月22日8月22日周六

星期六 · 2 条
01:58
Claude：Blog（网页）精选
AI 评分 61/100
Claude Mythos 5 网络安全能力扩展至更多防御者

Anthropic 宣布 Claude Mythos 5 现已集成至 Claude Security，并即将登陆合作伙伴的网络安全防御工具。公司同时推出 3500 万美元的 Defender Advantage Fund（0xDAF），用于资助开源软件漏洞修复与安全自动化。

Anthropic产品更新开源生态
另有 1 家信源报道MarkTechPost（RSS）
推荐理由：相较直接开放模型，以限定输出物提供补丁和告警的控制方式，为防御能力安全放量给出了可参照的产品分发思路，也解释了后续基金与验证计划的逻辑。
01:56
Ant Ling@AntLingAGI精选
AI 评分 53/100
蚂蚁百灵为SGLang推出权重缓存守护进程Today, we’re introducing the Weight Cache Daemon for SGLang. 🚀 On Ling-2.6-1T FP8, it reduced weight loading to ~0.63s, up to ~780× faster than disk loading, and cut total engine startup from 8.8 minutes to ~0.53 minutes. Here’s how it works.译今天，我们为 SGLang 推出 Weight Cache Daemon。🚀 在 Ling-2.6-1T FP8 上，它将权重加载时间缩短至约 0.63 秒，比磁盘加载快约 780 倍，并将引擎总启动时间从 8.8 分钟缩短至约 0.53 分钟。其工作原理如下。
产品更新开源生态部署/工程
另有 1 家信源报道LMSYS：Blog（Chatbot Arena 团队）
推荐理由：把权重加载从磁盘瓶颈改为缓存复用，对需要频繁扩缩容或重启推理服务的团队，启动耗时从分钟级降到秒级可能改变运维节奏。

8月21日8月21日周五

星期五 · 1 条
21:07
OpenBMB@OpenBMB精选
AI 评分 69/100
面壁智能 OpenBMB 推出 MathForm，面向 Lean 4 数学自动形式化的开源框架、数据集与模型🧮 Introducing MathForm, an open-source framework, dataset, and model for mathematical autoformalization with Lean 4.Formalizing mathematics makes mathematical knowledge machine-checkable, but it is more than translating statements into code. A model must map each concept onto the right types and definitions in Mathlib. A formal statement can compile and still misstate the original problem.Highlights ✨ MathForm Framework: Retrieval-Augmented, Verification-Guided Data Construction A retrieval planner pulls the Mathlib definitions and existing formalizations a statement needs. The generator then revises its output against Lean compiler diagnostics and semantic-consistency feedback for up to 3 rounds.FormalVerse Dataset: 367K+ Verified Lean 4 Examples Each example pairs a natural-language statement with verified Lean 4 code , across diverse mathematical domains and sources.Results 📊 • At a matched 100K budget with the same recipe and init, models trained on FormalVerse reach 60.32% Consistency Check, vs 46.53% on FineLeanCorpus and 41.49% on NuminaMath-LEAN • MathForm-8B achieves 88.06% Syntax Check and 72.37% Consistency Check Pass@8 across six benchmarks, outperforming ReForm-32B and Goedel-Formalizer-V2-32B at a quarter the size • On the hardest FATE-H / FATE-X subsets it reaches 63% / 37% Consistency Check, beating the strongest specialized baseline by 10 and 12 points🔗 Resources 📄 Paper: http://arxiv.org/abs/2608.14221 📚 Dataset: http://huggingface.co/datasets/openbmb/FormalVerse 🤖 Model: http://huggingface.co/openbmb/MathForm-8B 💻 Code: http://github.com/openbmb/MathForm译面壁智能 OpenBMB 推出 MathForm，一个面向 Lean 4 数学自动形式化的开源框架、数据集与模型。其 FormalVerse 数据集含 367K+ 已验证示例；在匹配 100K 预算下，基于其训练的模型 Consistency Check 达 60.32%，优于 FineLeanCorpus（46.53%）与 NuminaMath-LEAN（41.49%）。
开源生态推理数据/训练模型发布
推荐理由：数学形式化的难点是编译通过但语义漂移，该框架用检索规划和语义一致性反馈迭代修正，并开源数据集和 8B 模型，给小模型做形式化提供了可复用路径。

8月20日8月20日周四

星期四 · 2 条
10:00
公众号：蚂蚁百灵（Ling）精选
AI 评分 68/100
Ling-3.0 tiny & flash Base Models 正式开源

蚂蚁百灵开源 Ling-3.0-tiny-base 与 Ling-3.0-flash-base，并同步开放预训练、中期训练及 WSM 合并等共 6 个 checkpoints。tiny-base 总参数 7.9B、激活参数 1.3B，flash-base 总参数 124B、激活参数 5.1B，均采用统一训练方案，适合持续预训练、监督微调、偏好优化及强化学习后训练等场景。

开源生态数据/训练模型发布
另有 2 家信源报道X：蚂蚁百灵 (@AntLingAGI)IT之家（RSS）
推荐理由：Ling-3.0 开放中间 checkpoint 和 WSM 合并策略，研究者可比较各训练阶段增益，并在自己数据上从合适节点继续预训练或后训练，而不只是拿到最终权重。
05:04
Hao AI Lab@haoailab精选
AI 评分 71/100
FastMetal 让 Mac 本地 30 秒生成视频A 5-second 480P video, generated entirely on a Mac, in 30 seconds. No CUDA, no cloud, 3.9 GiB of memory.FastMetal brings the FastWan-QAD family to Apple Silicon. The DiT, the DMD sampler, and the decoder all run on Metal through MLX, INT8 by default.Three models: 1.3B at 480P, 5B at 720P, 14B for quality.📷 Blog: http://haoailab.com/blogs/fastmetal 📷 Code: http://github.com/hao-ai-lab/FastVideo 📷 Models: http://huggingface.co/collections/FastVideo/fastmetal译一段 5 秒 480P 视频，完全在 Mac 上生成，耗时 30 秒。无需 CUDA，无需云端，仅占用 3.9 GiB 内存。 FastMetal 将 FastWan-QAD 系列带到 Apple Silicon。DiT、DMD 采样器和解码器均通过 MLX 在 Metal 上运行，默认 INT8。 三个模型：1.3B 支持 480P，5B 支持 720P，14B 追求画质。 📷 博客：http://haoailab.com/blogs/fastmetal 📷 代码：http://github.com/hao-ai-lab/FastVideo 📷 模型：http://huggingface.co/collections/FastVideo/fastmetal
Hugging Face开源/仓库端侧视频
推荐理由：在Apple Silicon本地生成视频，把云端视频生成的计算依赖降到了消费级硬件，为需要数据隐私或低成本原型创作的工作流提供了替代路径。

8月19日8月19日周三

星期三 · 3 条
22:17
公众号：百度智能云（文心）精选
AI 评分 64/100
百度千帆上架GLM-5.3

百度千帆今日上架智谱开源旗舰模型GLM-5.3，API定价与智谱官方保持一致。该模型与上代同架构、同参数，依靠后训练Scaling在代码与网络安全能力上表现超预期，AA综合智能指数取得60分，与Kimi K3并列开源模型第一。开发者可通过API调用或订阅Token Plan企业版接入Claude Code等AI工具使用。

产品更新开源生态编码
推荐理由：GLM-5.3 与上代同参数却靠后训练 Scaling 进入前沿，单任务成本又是同档最低，这让原本受调用成本限制的长链路智能体和编码任务有了开源可选项。
09:11
公众号：智谱（GLM）精选
AI 评分 79/100
GLM-5.3上线：AA智能指数60分并列开源第一，成本更低

GLM-5.3 API即日上线，擅长复杂编码、防御性网络安全与长程任务，在AA综合智能指数中取得60分，与Claude Fable 5、GPT-5.6 Sol等闭源旗舰同级，并与Kimi K3并列开源模型第一。该模型以更小参数规模和更低调用成本降低前沿智能门槛，单任务成本为旗舰模型中最低。API定价与GLM-5.2持平，模型权重将于下周五开源。

产品更新开源生态推理编码
另有 5 家信源报道IT之家（RSS）X：X.PIN (@thexpin)The Decoder：AI News（RSS）X：Artificial Analysis (@ArtificialAnlys)X：智谱 Z.ai (@Zai_org)
推荐理由：GLM-5.3 把前沿智能的单任务成本压到同档最低，模型选型时成本与智能可以放在同一优先级比较，而不必默认高价闭源旗舰。
05:54
Hacker News 热门（buzzing.cc 中文翻译）精选
AI 评分 74/100
Mojo 语言正式开源，编译器与工具链全面开放

Mojo🔥 语言现已正式开源，采用 Apache 2.0 许可证（含 LLVM 例外），编译器、工具链及全部源码已发布至 modular GitHub 仓库。Mojo 上周刚达成 1.0 版本（源码稳定），此次开源涵盖整个编译器与工具链。目前暂不接受编译器相关贡献，计划年底前开放，标准库自 2024 年起已接受社区贡献。

开源/仓库开源生态部署/工程
另有 3 家信源报道IT之家（RSS）Hacker News 热门（buzzing.cc 中文翻译）Simon Willison 博客
推荐理由：Mojo 从闭源编译器转为可自行构建的工具链，开发者能直接查看和修改标准库实现，减少对厂商二进制分发的单一依赖。

8月18日8月18日周二

星期二 · 3 条
22:05
Hugging Face：Blog（RSS）精选
AI 评分 75/100
Sentence Transformers v6.0 新增 MultiVectorEncoder，支持 ColBERT 风格多向量模型

Sentence Transformers v6.0 新增第四种模型类型 MultiVectorEncoder，可直接加载 PyLate、Stanford-NLP ColBERT 及 colpali-engine 检查点，用于 ColBERT 式晚期交互检索。

产品更新开源生态搜索
推荐理由：与同骨干的稠密模型相比，多向量检索在平均 NDCG 上高出约一个点，代价是索引体积增大数十倍，适合需要保留逐词精确匹配的场景。
15:36
Google AI：DEV 作者专属（RSS）精选
AI 评分 67/100
设计 AI 评测：先求清晰，再谈可视化

本文演示如何用开源评测框架 Inspect AI 和 Harbor 评估 agent 技能，并借助 Google Sheets 和 Data Studio 进行可视化分析。

智能体Google开源生态教程/实践
推荐理由：把技能评估拆成模型基线、技能条件与多维事实打分，让团队用同一套脚本量化技能增量，而不是凭单次演示做判断。
13:34
蚂蚁 inclusionAI：GitHub 新仓库精选
AI 评分 64/100
inclusionAI 开源 ConceptEdit：基于概念缩放与密集监督的图像编辑数据生成管线

蚂蚁集团 inclusionAI 开源 ConceptEdit，一个基于概念缩放与密集监督的图像编辑数据生成管线。该管线通过三阶段流程（VLM 生成指令、FLUX 执行编辑、VQA 评估筛选）构建大规模、基于分类法的图像编辑数据集，并提供单概念与多概念两种变体。项目采用 MIT 许可证，支持断点续跑，需 OpenAI 兼容 VLM 端点与本地 FLUX 检查点。

GitHub图像生成开源/仓库数据/训练
推荐理由：流水线把图像编辑数据生成拆为指令生成、FLUX 编辑、VLM 评判三步，多概念版本将多个并行编辑合并为一条指令并支持断点续跑，为构建带质检的编辑训练数据提供可复用框架。

8月15日8月15日周六

星期六 · 2 条
08:00
HuggingFace Daily Papers（社区热门论文）精选
AI 评分 72/100
MOSS-VL技术报告：将实时交互作为一等能力的开源视觉语言模型家族

MOSS-VL是一个将实时交互（边感知边说话）作为一等能力的开源视觉语言模型家族，通过门控交叉注意力让语言解码器在生成时同步处理视觉输入。MOSS-VL-Realtime在四个流式基准中平均成绩居开源模型之首（三项第一、一项第二），在OmniMMI Proactive Alerting上以66.0分大幅领先最佳基线（37.5分）。

多模态开源生态模型发布视频
推荐理由：把视觉 token 放在解码序列之外，并用门控交叉注意力让模型边生成边接收画面，给低延迟多模态交互提供了一个具体架构参考。
00:03
Hugging Face：Blog（RSS）精选
AI 评分 71/100
2026年夏季开源模型生态观察：中国前沿模型规模领先，AMD与NVIDIA主导发布量

2026年1至8月，Hugging Face公开模型仓库从243万增至296万，但85.6%的模型下载量不足200次，1.5%的仓库占据99.2%下载量。中国实验室月度最大开源模型参数规模在754B至2.78万亿之间，美国实验室七个月中五个月低于130B。AMD与NVIDIA各发布超200个新模型仓库，成为发布开源模型最多的机构。

Hugging Face开源生态现象/趋势
推荐理由：下载量与点赞量分别记录实际依赖和社区兴奋点，把两者混为同一个热度指标是评估开源模型生态时最常见的偏差。

8月14日8月14日周五

星期五 · 2 条
23:22
Z.ai@Zai_org精选
AI 评分 66/100
GLM-5.3 开放前安全评估：网络防御能力显著提升https://x.com/i/article/2088262763327971328Preparing GLM-5.3 for Open Release: A Responsible Path to Cyber DefenseWhen GLM-5.2 helped Hugging Face investigate an incident in which an AI autonomously bypassed its own safeguards, it highlighted a broader shift. AI is becoming part of both cyber offense and cyber defense.As powerful cyber capabilities become more accessible, strong defensive capabilities cannot remain limited to a small number of well-resourced organizations. Open-source maintainers, independent researchers, developers, and smaller security teams also need tools that can help them find and fix vulnerabilities before they are exploited.An open world cannot have only open attack surfaces. It must also have an open shield.GLM-5.3 is our most capable model to date for cybersecurity tasks. It delivers substantial improvements in vulnerability discovery, exploit analysis, and complex multistep security tasks. These capabilities can help defenders identify weaknesses earlier, validate risks, and accelerate remediation.They also create clear dual-use risks. We are therefore taking a staged approach to release. Selected security partners will first evaluate GLM-5.3 in controlled settings. Broader access and API availability will follow. Once the necessary safety evaluations and release preparations are complete, we will publish GLM-5.3’s complete model weights.Responsible openness does not mean treating every capability as harmless. It means evaluating risks transparently, strengthening safeguards before release, coordinating the disclosure of validated vulnerabilities, and expanding access to advanced defensive capabilities in ways proportionate to the risks.From vulnerability discovery to multistep security analysisAs part of post-training, we introduced vulnerability discovery data and authorized security environments into the training mix. We expected this to improve the model’s ability to find and analyze vulnerabilities.As training scaled, the improvement extended beyond isolated flaws. GLM-5.3 became more effective at connecting vulnerability conditions, program behavior, validation paths, and potential impact across multiple stages of analysis.We evaluate these capabilities across three benchmarks:• CyberGym begins with white-box source code and tests whether a model can identify and validate vulnerabilities by triggering faults. GLM-5.3 scores 84.5%, compared with 77.2% for GLM-5.2.• ExploitBench requires deeper reasoning about real vulnerabilities and their exploitation. GLM-5.3 reaches 54.4%, more than twice GLM-5.2’s 24.4%.• ExploitGym measures completed exploitation tasks under normalized evaluation budgets. GLM-5.3 completes 105 tasks within two hours and 130 within six hours, compared with 29 and 39 for GLM-5.2.The pattern is consistent. GLM-5.3 improves most over GLM-5.2 as tasks move from isolated vulnerability discovery toward multistep exploitation. The results also show where further progress is needed, particularly on the most complex end-to-end tasks.From benchmarks to real softwareWe have also worked with universities and professional security teams to evaluate GLM models on real-world codebases in authorized settings.Across this work, the GLM series has produced 2,436 vulnerability findings across 269 projects, including 1,097 categorized as medium-to-high severity. These findings span system software, operating systems, browser engines, open-source infrastructure, web applications, network protocols, and intelligent devices. Some of the underlying issues had remained unnoticed for decades.In these evaluations, security experts establish the authorized scope, review model outputs, investigate potential risks, and coordinate with the relevant parties. GLM models can help researchers reconstruct complex program logic, narrow large numbers of candidate paths, and connect evidence across multiple components.The purpose is not simply to generate more findings. It is to help defenders identify meaningful risks earlier and reduce the time between discovery and remediation.Discovery must be followed by responsible disclosureA vulnerability is not safely handled at the moment it is discovered. It must be reviewed, reproduced where appropriate, reported through the proper channels, and coordinated with the affected maintainers.Findings from our security work are submitted through established disclosure processes. We publish technical details only when doing so is consistent with the relevant disclosure and remediation process. For issues that remain under coordination, we do not release information that could unnecessarily increase risk or identify affected projects.To make this work more transparent, we created the Z.ai Security Disclosure Ledger.The ledger records findings as they move through the disclosure process. For publicly disclosed issues, it may include the affected project, severity, a CVE or other identifier where available, and information about how long the issue remained in the codebase.For vulnerabilities still under coordinated disclosure, the ledger can publish a cryptographic hash. This allows a finding to be verified later without prematurely revealing operational details.Opening a model and disclosing a vulnerability are separate decisions. Making a model more broadly available does not require publishing vulnerability details before maintainers have had an appropriate opportunity to investigate and respond.Safety and staged releaseCybersecurity is a particularly difficult domain for AI safety. Offensive and defensive tasks often involve the same terminology, code, and technical methods.A request to analyze a vulnerability could come from a maintainer preparing a patch, a student solving a CTF challenge, a researcher conducting an authorized assessment, or an attacker targeting a real system. Keywords alone cannot reliably distinguish these cases. Intent, authorization, context, target, and potential impact all matter.For GLM-5.3, we use a defense-in-depth approach with three complementary layers.External classifierIn our hosted services, an external classifier identifies high-risk requests and helps prevent clearly harmful activity.Reasoning monitorA reasoning monitor assesses risk during task execution. It is designed to detect harmful objectives that may emerge across multiple steps rather than relying only on the wording of the initial request.Deep safety alignmentThe model itself is trained to distinguish legitimate security work from high-risk offensive activity and to refuse requests that cross that boundary.Deep safety alignment is particularly important for an open-weight release. Hosted classifiers and monitors apply to our services, but they do not automatically accompany the model into every local deployment. Model-level alignment is the safety layer included in the released checkpoint.To develop these systems, we created differential training data that reflects both the similarities and the differences between authorized security research and malicious activity. We also constructed adversarial data covering jailbreak variants, disguised intent, and other attempts to evade safety review.Our evaluations cover a range of cybersecurity tasks, including:• security education and knowledge;• blue-team defense;• CTF challenges;• vulnerability discovery and remediation;• authorized penetration testing;• exploit development;• unauthorized intrusion and other clearly malicious activity.The objective is to reduce high-risk abuse without broadly refusing legitimate defensive, educational, and research tasks.Before broader release, professional security teams will conduct safety evaluations and red-team testing. These evaluations examine both whether the model can be manipulated into supporting harmful activity and whether its safeguards interfere with legitimate security work.No safety system can eliminate every dual-use risk. Once model weights are public, no developer can guarantee control over every downstream modification or use. Model-level safeguards can raise the barrier to abuse, but they cannot provide absolute control.Our release process therefore focuses on the stages where meaningful risk reduction is possible: training, pre-release evaluation, controlled partner testing, hosted-service safeguards, responsible disclosure, and continuing adversarial testing.Launching the OpenVuln initiativeMuch of the world’s digital infrastructure depends on open-source software. Many critical projects are maintained by small teams or individual contributors without dedicated security resources.At the same time, AI is making complex cyber tasks easier to automate. If advanced defensive capabilities remain concentrated within a small number of organizations, the projects with the fewest resources may be left protecting some of the most important parts of the software supply chain.To help address this imbalance, we are launching the OpenVuln initiative alongside GLM-5.3.Continuous support for open-source securityWe will work with maintainers to audit important open-source projects, identify potential vulnerabilities, and support responsible disclosure and remediation.Maintainers can use OpenVuln to submit projects for security review and learn more about the process.A shield for the open worldGLM-5.3 shows that open models can become meaningfully stronger at vulnerability discovery, exploit analysis, and complex security reasoning. That progress carries real defensive value and real dual-use risk.Our responsibility is to direct these capabilities toward finding vulnerabilities earlier, supporting responsible remediation, and strengthening the open-source systems on which everyone depends.Following staged evaluation and broader API access, we intend to release GLM-5.3 as an open-weight model. We will continue improving model-level safeguards, testing adversarial use, and supporting coordinated disclosure throughout that process.The open world must have a shield of its own. Through GLM-5.3 and the OpenVuln initiative, we intend to make that shield more broadly available and to release it with care.译智谱发布 GLM-5.3，为网络安全任务最强模型，在漏洞发现、利用分析和多步安全任务上大幅提升。CyberGym 得分 84.5%（GLM-5.2 为 77.2%），ExploitBench 达 54.4%（此前 24.4%），ExploitGym 两小时完成 105 项任务（此前 29 项）。因双重用途风险，将先由安全伙伴受控评估，再开放 API 和完整权重。
安全/对齐开源生态模型发布
另有 1 家信源报道X：Testing Catalog (@testingcatalog)
推荐理由：比单点漏洞发现更值得关注的是 GLM-5.3 在多步利用任务上的提升，这细分了安全团队能够交给模型的工作类型，从定位缺陷到验证攻击路径。
23:22
Qwen@Alibaba_Qwen精选
AI 评分 71/100
通义千问开源 Qwen3.8 系列模型We promised open weights for Qwen3.8. Now, time to meet them! 🎉⚡ Qwen3.8-27B: • A native multimodal dense model. With just 27B parameters, it outperforms Qwen3.7-Plus overall and shines in real-world coding & office workflows. • 262K native context, easily extendable to 1M tokens via YaRN. • Built for builders. Highly efficient, high-quality, and licensed under Apache 2.0.🚀 The open weights for Qwen3.8-2.4T-A95B (Max-level) have also been released recently.Whether you're shipping lightweight applications with Qwen3.8-27B locally or building agents with Qwen3.8-2.4T-A95B, they're yours now!Download, deploy, and build something we haven't imagined yet. 👀👇 • Hugging Face： https://huggingface.co/collections/Qwen/qwen38 • ModelScope： https://www.modelscope.cn/collections/Qwen/Qwen38译通义千问兑现承诺，开源 Qwen3.8 系列模型。其中 Qwen3.8-27B 为原生多模态稠密模型，仅 27B 参数即全面超越 Qwen3.7-Plus，原生支持 262K 上下文，可通过 YaRN 扩展至 1M tokens，采用 Apache 2.0 许可。Max 级 Qwen3.8-2.4T-A95B 的开放权重也已同步发布。
多模态开源生态模型发布
另有 10 家信源报道Simon Willison 博客X：Testing Catalog (@testingcatalog)X：通义千问 / Qwen (@Alibaba_Qwen)Hacker News 热门（buzzing.cc 中文翻译）The Decoder：AI News（RSS）X：阿里云 / Alibaba Cloud (@alibaba_cloud)X：Rohan Paul (@rohanpaul_ai)IT之家（RSS）X：阿易 AI Notes (@AYi_AInotes)X：Kim (@kimmonismus)
推荐理由：官方称 27B 稠密多模态模型整体表现超过 Qwen3.7-Plus，262K 原生上下文可扩至 1M，Apache 2.0 协议下本地部署轻量应用多了一个更高效的选择。
模型发布Hugging FaceAI 编码