{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-07-23T06:40:50.084Z","headline":"思科推出 Antares 小语言模型，专用于漏洞定位","description":"思科发布 Antares 系列小语言模型，专门用于漏洞定位，已在 Hugging Face 发布 350M 和 1B 参数版本，后续将推出 3B 版本。在思科的漏洞定位基准测试中，Antares-1B 的漏洞文件检出率达 22.4%，略高于 GPT-5.5 的 22.1%。该模型可本地部署，但仅作辅助用途，企业仍需结合其他安全措施建立完整防护体系。","url":"https://www.aioga.com/news/cmrvynans02ekbipza1gsp8tb/","mainEntityOfPage":"https://www.aioga.com/news/cmrvynans02ekbipza1gsp8tb/","datePublished":"2026-07-22T10:03:39.000Z","dateModified":"2026-07-22T10:03:39.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://www.ithome.com/0/980/239.htm","https://aihot.virxact.com/items/cmrvynans02ekbipza1gsp8tb"],"canonicalUrl":"https://www.aioga.com/news/cmrvynans02ekbipza1gsp8tb/","directAnswer":{"@type":"Answer","text":"Aioga 编辑摘要：思科发布 Antares 系列小语言模型，专门用于漏洞定位，已在 Hugging Face 发布 350M 和 1B 参数版本，后续将推出 3B 版本。 Aioga 将其归入「模型更新」方向，重点关注它对真实使用和行业竞争的影响。","url":"https://www.aioga.com/news/cmrvynans02ekbipza1gsp8tb/","dateCreated":"2026-07-22T10:03:39.000Z","author":{"@type":"Organization","@id":"https://www.aioga.com/authors/aioga-editorial/#editorial-team","name":"Aioga Editorial Team","url":"https://www.aioga.com/authors/aioga-editorial/"}},"evidence":[{"@type":"CreativeWork","name":"IT之家 source article","url":"https://www.ithome.com/0/980/239.htm","datePublished":"2026-07-22T10:03:39.000Z","provider":{"@type":"Organization","name":"IT之家","url":"https://www.ithome.com/0/980/239.htm"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.virxact.com/items/cmrvynans02ekbipza1gsp8tb","datePublished":"2026-07-22T10:03:39.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.virxact.com/items/cmrvynans02ekbipza1gsp8tb"}}],"aggregationSource":"IT之家（RSS）","originalPublisher":{"name":"IT之家","url":"https://www.ithome.com/0/980/239.htm"},"article":{"id":"cmrvynans02ekbipza1gsp8tb","slug":"cmrvynans02ekbipza1gsp8tb","url":"https://www.aioga.com/news/cmrvynans02ekbipza1gsp8tb/","title":"思科推出 Antares 小语言模型，专用于漏洞定位","title_en":"思科推出 Antares 小语言 AI 模型，专为\"定位漏洞\"设计","summary":"思科发布 Antares 系列小语言模型，专门用于漏洞定位，已在 Hugging Face 发布 350M 和 1B 参数版本，后续将推出 3B 版本。在思科的漏洞定位基准测试中，Antares-1B 的漏洞文件检出率达 22.4%，略高于 GPT-5.5 的 22.1%。该模型可本地部署，但仅作辅助用途，企业仍需结合其他安全措施建立完整防护体系。","source":"IT之家（RSS）","sourceUrl":"https://www.ithome.com/0/980/239.htm","aiHotUrl":"https://aihot.virxact.com/items/cmrvynans02ekbipza1gsp8tb","publishedAt":"2026-07-22T10:03:39.000Z","category":"模型更新","score":59,"selected":false,"articleBody":["IT之家：https://www.ithome.com/ 7 月 22 日消息，思科发文，宣布推出 Antares 系列小语言模型，相应模型专门用于定位漏洞，目前官方已在 Hugging Face（点此访问：https://huggingface.co/collections/fdtn-ai/antares）发布 350M 和 1B 参数版本，后续还将推出 30 亿参数（3B）版本。","据介绍，Antares 并非能够自主发现未知漏洞的 AI 模型，其设计目的是帮助安全人员进行漏洞定位（Vulnerability Localization），模型在获取 CWE 漏洞类型及相关描述后，会利用终端命令搜索代码库，输出检查报告，以便帮助安全人员提升人工审查效率。","为了评估 Antares 模型性能，思科进行了一项漏洞定位基准测试（Vulnerability Localization Benchmark），该测试涵盖 290 个真实代码库、147 种 CWE 漏洞类型以及 500 项测试任务。根据思科公布的技术报告，Antares-1B 在漏洞文件检出率（Recall）方面达到 22.4%， 在所有参与测试的模型中排名最高，略高于 GPT-5.5 的 22.1% 。","思科表示，由于 Antares 模型规模较小，可以直接部署在本地环境、企业内网中运行，具有轻量化、隐私化特点。 不过其本身功能有限，只能用作辅助用途 ，企业仍需要结合依赖组件分析、软件组成分析（SBOM）、敏感信息扫描、动态安全测试、威胁建模、漏洞修复以及人工安全审查等措施，才能建立完整的软件安全防护体系。"],"articleImages":[{"sourceUrl":"https://img.ithome.com/newsuploadfiles/2026/7/8b79b68a-976a-4241-bc78-ef88ac07a052.jpg?x-bce-process=image/format,f_auto","alt":"","afterParagraph":0,"url":"/media/articles/cmrvynans02ekbipza1gsp8tb/80e5148dbd42e7f0.jpg"},{"sourceUrl":"https://img.ithome.com/newsuploadfiles/2026/7/28dafeb9-8a6a-4683-af48-f551236b17c8.png?x-bce-process=image/format,f_auto","alt":"","afterParagraph":2,"url":"/media/articles/cmrvynans02ekbipza1gsp8tb/562683acbd098a7b.png"},{"sourceUrl":"https://img.ithome.com/newsuploadfiles/2026/7/24556621-d866-43f6-a0f7-b2d080b09edf.png?x-bce-process=image/format,f_auto","alt":"","afterParagraph":3,"url":"/media/articles/cmrvynans02ekbipza1gsp8tb/0a54cc2b33f16d59.png"}],"mediaStatus":"ok","articleBodyZh":[],"translationStatus":"","bodyOrigin":"source-page","editorial":{"summary":"Aioga 编辑摘要：思科发布 Antares 系列小语言模型，专门用于漏洞定位，已在 Hugging Face 发布 350M 和 1B 参数版本，后续将推出 3B 版本。 Aioga 将其归入「模型更新」方向，重点关注它对真实使用和行业竞争的影响。","background":"背景分析：模型与研究类动态需要结合能力边界、开放方式、成本、可用性和真实任务表现判断，单项指标领先不等于已经形成稳定采用。","viewpoint":"Aioga 判断：这条动态更适合作为行业观察信号，当前信息足以建立线索，但不足以推导长期结论。","implications":"影响分析：对相关团队而言，短期应先核对来源、可用范围和实际成本，再判断是否值得接入或跟进。","nextStep":"后续观察：继续观察官方文档、实际可用性、价格变化、开发者反馈和竞品回应。","evidenceRefs":["title","summary","articleBody"],"confidence":"medium","status":"published","aiGenerated":false,"autoApproved":true,"generatedBy":"rule-safe-fallback","generatedAt":"2026-07-23T06:49:19.035Z","sourceHash":"cb84fa3c485b6964","validation":{"passed":true,"mode":"rule-safe-fallback","checks":["schema","length","source-attribution","no-html"]}},"tags":["模型更新","IT之家（RSS）"],"translations":{"zh-CN":{"title":"思科推出 Antares 小语言模型，专用于漏洞定位","summary":"思科发布 Antares 系列小语言模型，专门用于漏洞定位，已在 Hugging Face 发布 350M 和 1B 参数版本，后续将推出 3B 版本。在思科的漏洞定位基准测试中，Antares-1B 的漏洞文件检出率达 22.4%，略高于 GPT-5.5 的 22.1%。该模型可本地部署，但仅作辅助用途，企业仍需结合其他安全措施建立完整防护体系。","category":"模型更新","source":"IT之家","aggregationSource":"IT之家（RSS）","pageTitle":"思科推出 Antares 小语言模型，专用于漏洞定位 - Aioga AI资讯","description":"思科发布 Antares 系列小语言模型，专门用于漏洞定位，已在 Hugging Face 发布 350M 和 1B 参数版本，后续将推出 3B 版本。在思科的漏洞定位基准测试中，Antares-1B 的漏洞文件检出率达 22.4%，略高于 GPT-5.5 的 22.1%。该模型可本地部署，但仅作辅助用途，企业仍需结合其他安全措施建立完整防护体系。","url":"https://www.aioga.com/news/cmrvynans02ekbipza1gsp8tb/"},"en":{"title":"Cisco launches Antares small language model, specifically for vulnerability localization","summary":"Cisco has released the Antares series of small language models, specifically for vulnerability detection, with 350M and 1B parameter versions already published on Hugging Face, and a 3B version to be released later. In Cisco's vulnerability detection benchmark tests, Antares-1B achieved a 22.4% detection rate for vulnerable files, slightly higher than GPT-5.5's 22.1%. This model can be deployed locally, but is intended only for auxiliary use, and enterprises still need to combine it with other security measures to establish a complete protection system.","category":"Models","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Cisco launches Antares small language model, specifically for vulnerability localization - Aioga AI News","description":"Cisco has released the Antares series of small language models, specifically for vulnerability detection, with 350M and 1B parameter versions already published on Hugging Face, and...","url":"https://www.aioga.com/en/news/cmrvynans02ekbipza1gsp8tb/","contentTranslated":true,"sourceHash":"ef2a104326f431fe","translatedAt":"2026-07-22T23:02:16.747Z"},"ja":{"title":"シスコ、脆弱性の特定専用の小型言語モデル Antares を発表","summary":"シスコは Antares シリーズの小規模言語モデルを発表しました。これは特に脆弱性の特定に使用され、Hugging Face で 350M と 1B パラメータ版が公開されており、今後 3B 版もリリース予定です。シスコの脆弱性特定ベンチマークテストでは、Antares-1B の脆弱性ファイル検出率は 22.4% で、GPT-5.5 の 22.1% をわずかに上回っています。このモデルはローカルでの導入が可能ですが、あくまで補助用途に限られ、企業は他のセキュリティ対策と組み合わせて完全な防御体制を構築する必要があります。","category":"モデル更新","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"シスコ、脆弱性の特定専用の小型言語モデル Antares を発表 - Aioga AIニュース","description":"シスコは Antares シリーズの小規模言語モデルを発表しました。これは特に脆弱性の特定に使用され、Hugging Face で 350M と 1B パラメータ版が公開されており、今後 3B 版もリリース予定です。シスコの脆弱性特定ベンチマークテストでは、Antares-1B の脆弱性ファイル検出率は 22.4% で、GPT-5.5 の 22.1% をわず...","url":"https://www.aioga.com/ja/news/cmrvynans02ekbipza1gsp8tb/","contentTranslated":true,"sourceHash":"ef2a104326f431fe","translatedAt":"2026-07-22T23:02:28.447Z"},"ko":{"title":"시스코, 취약점 위치 지정 전용 Antares 소형 언어 모델 출시","summary":"시스코가 Antares 시리즈 소형 언어 모델을 발표했으며, 이는 취약점 위치 확인 전용으로, Hugging Face에 350M 및 1B 파라미터 버전이 공개되었고, 추후 3B 버전을 출시할 예정입니다. 시스코의 취약점 위치 벤치마크 테스트에서 Antares-1B의 취약 파일 검출률은 22.4%로, GPT-5.5의 22.1%보다 약간 높습니다. 이 모델은 로컬에 배포할 수 있으나, 보조 용도로만 사용되며, 기업은 여전히 다른 보안 조치와 결합하여 완전한 방어 체계를 구축해야 합니다.","category":"모델 업데이트","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"시스코, 취약점 위치 지정 전용 Antares 소형 언어 모델 출시 - Aioga AI 뉴스","description":"시스코가 Antares 시리즈 소형 언어 모델을 발표했으며, 이는 취약점 위치 확인 전용으로, Hugging Face에 350M 및 1B 파라미터 버전이 공개되었고, 추후 3B 버전을 출시할 예정입니다. 시스코의 취약점 위치 벤치마크 테스트에서 Antares-1B의 취약 파일 검출률은 22.4%로, GPT-5.5의 22....","url":"https://www.aioga.com/ko/news/cmrvynans02ekbipza1gsp8tb/","contentTranslated":true,"sourceHash":"ef2a104326f431fe","translatedAt":"2026-07-22T23:03:11.885Z"},"es":{"title":"Cisco lanza el pequeño modelo de lenguaje Antares, especialmente diseñado para la localización de vulnerabilidades","summary":"Cisco ha lanzado la serie de modelos de lenguaje pequeños Antares, específicamente diseñados para la localización de vulnerabilidades, y ya ha publicado versiones de 350M y 1B parámetros en Hugging Face, con una versión de 3B planeada para el futuro. En las pruebas de referencia de localización de vulnerabilidades de Cisco, la tasa de detección de archivos con vulnerabilidades de Antares-1B alcanzó el 22.4%, ligeramente superior al 22.1% de GPT-5.5. Este modelo se puede desplegar localmente, pero solo como apoyo; las empresas aún necesitan combinarlo con otras medidas de seguridad para establecer un sistema de protección completo.","category":"Modelos","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Cisco lanza el pequeño modelo de lenguaje Antares, especialmente diseñado para la localización de vulnerabilidades - Aioga Noticias de IA","description":"Cisco ha lanzado la serie de modelos de lenguaje pequeños Antares, específicamente diseñados para la localización de vulnerabilidades, y ya ha publicado versiones de 350M y 1B pará...","url":"https://www.aioga.com/es/news/cmrvynans02ekbipza1gsp8tb/","contentTranslated":true,"sourceHash":"ef2a104326f431fe","translatedAt":"2026-07-22T23:03:07.624Z"},"fr":{"title":"Cisco lance le petit modèle de langage Antares, spécialement conçu pour la localisation des vulnérabilités","summary":"Cisco a publié la série de petits modèles de langage Antares, spécialement conçue pour la localisation des vulnérabilités, avec des versions de 350M et 1B paramètres déjà disponibles sur Hugging Face, et une version 3B prévue ultérieurement. Dans le test de référence de localisation des vulnérabilités de Cisco, le taux de détection des fichiers vulnérables d'Antares-1B atteint 22,4 %, légèrement supérieur aux 22,1 % de GPT-5.5. Ce modèle peut être déployé localement, mais uniquement à des fins d'assistance ; les entreprises doivent toujours combiner d'autres mesures de sécurité pour établir un système de protection complet.","category":"Modèles","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Cisco lance le petit modèle de langage Antares, spécialement conçu pour la localisation des vulnérabilités - Aioga Actualités IA","description":"Cisco a publié la série de petits modèles de langage Antares, spécialement conçue pour la localisation des vulnérabilités, avec des versions de 350M et 1B paramètres déjà disponibl...","url":"https://www.aioga.com/fr/news/cmrvynans02ekbipza1gsp8tb/","contentTranslated":true,"sourceHash":"ef2a104326f431fe","translatedAt":"2026-07-22T23:03:56.392Z"},"de":{"title":"Cisco bringt das Antares-Kleinsprachenmodell auf den Markt, das speziell für die Schwachstellenortung entwickelt wurde","summary":"Cisco hat die Antares-Serie kleiner Sprachmodelle veröffentlicht, die speziell für die Schwachstellenerkennung entwickelt wurden. Auf Hugging Face sind bereits Versionen mit 350M und 1B Parametern verfügbar, eine 3B-Version wird später erscheinen. In Ciscos Benchmark-Test zur Schwachstellenerkennung erreichte Antares-1B eine Erkennungsrate von 22,4 % bei Schwachstellendateien, etwas höher als die 22,1 % von GPT-5.5. Das Modell kann lokal eingesetzt werden, dient jedoch nur als Hilfsmittel; Unternehmen müssen weiterhin andere Sicherheitsmaßnahmen kombinieren, um ein vollständiges Schutzsystem aufzubauen.","category":"模型更新","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Cisco bringt das Antares-Kleinsprachenmodell auf den Markt, das speziell für die Schwachstellenortung entwickelt wurde - Aioga KI-News","description":"Cisco hat die Antares-Serie kleiner Sprachmodelle veröffentlicht, die speziell für die Schwachstellenerkennung entwickelt wurden. Auf Hugging Face sind bereits Versionen mit 350M u...","url":"https://www.aioga.com/de/news/cmrvynans02ekbipza1gsp8tb/","contentTranslated":true,"sourceHash":"ef2a104326f431fe","translatedAt":"2026-07-22T23:03:51.870Z"},"pt-BR":{"title":"A Cisco lançou o pequeno modelo de linguagem Antares, destinado especificamente à localização de vulnerabilidades","summary":"A Cisco lançou a série de pequenos modelos de linguagem Antares, especificamente para localização de vulnerabilidades, já disponibilizando as versões de 350M e 1B parâmetros no Hugging Face, e futuramente lançará a versão de 3B. Nos testes de benchmark de localização de vulnerabilidades da Cisco, a taxa de detecção de arquivos vulneráveis do Antares-1B atingiu 22,4%, ligeiramente superior aos 22,1% do GPT-5.5. Este modelo pode ser implantado localmente, mas apenas como um auxílio, sendo que as empresas ainda precisam combinar outras medidas de segurança para estabelecer um sistema de proteção completo.","category":"模型更新","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"A Cisco lançou o pequeno modelo de linguagem Antares, destinado especificamente à localização de vulnerabilidades - Aioga Notícias de IA","description":"A Cisco lançou a série de pequenos modelos de linguagem Antares, especificamente para localização de vulnerabilidades, já disponibilizando as versões de 350M e 1B parâmetros no Hug...","url":"https://www.aioga.com/pt-BR/news/cmrvynans02ekbipza1gsp8tb/","contentTranslated":true,"sourceHash":"ef2a104326f431fe","translatedAt":"2026-07-22T23:04:35.206Z"},"ru":{"title":"Cisco выпустила маленькую языковую модель Antares, специально для определения уязвимостей","summary":"Cisco выпустила серию небольших языковых моделей Antares, специально предназначенных для локализации уязвимостей. На Hugging Face уже доступны версии с 350 млн и 1 млрд параметров, в дальнейшем будет выпущена версия с 3 млрд параметров. В бенчмарках Cisco по локализации уязвимостей Antares-1B показала уровень обнаружения уязвимостей в файлах 22,4%, немного выше, чем у GPT-5.5, у которой 22,1%. Эта модель может быть развернута локально, но только для вспомогательных целей, предприятиям по-прежнему необходимо сочетать её с другими мерами безопасности для построения полноценной системы защиты.","category":"模型更新","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Cisco выпустила маленькую языковую модель Antares, специально для определения уязвимостей - Aioga Новости ИИ","description":"Cisco выпустила серию небольших языковых моделей Antares, специально предназначенных для локализации уязвимостей. На Hugging Face уже доступны версии с 350 млн и 1 млрд параметров,...","url":"https://www.aioga.com/ru/news/cmrvynans02ekbipza1gsp8tb/","contentTranslated":true,"sourceHash":"ef2a104326f431fe","translatedAt":"2026-07-22T23:04:36.043Z"},"ar":{"title":"كيسكو تطلق نموذج اللغة الصغيرة Antares، مخصص لتحديد الثغرات","summary":"أصدرت شركة سيسكو سلسلة نماذج اللغة الصغيرة Antares، المخصصة لتحديد الثغرات، وقد تم إصدار إصدارات ذات 350 مليون و1 مليار معلمة على Hugging Face، وسيتم لاحقًا إطلاق إصدار 3 مليارات معلمة. في اختبارات القياس المعياري لتحديد الثغرات لدى سيسكو، بلغت نسبة اكتشاف ملفات الثغرات لـ Antares-1B حوالي 22.4٪، وهي أعلى قليلاً من نسبة GPT-5.5 البالغة 22.1٪. يمكن نشر هذا النموذج محليًا، لكنه يُستخدم للأغراض المساعدة فقط، ولا تزال الشركات بحاجة إلى دمجه مع تدابير أمان أخرى لإنشاء نظام حماية كامل.","category":"模型更新","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"كيسكو تطلق نموذج اللغة الصغيرة Antares، مخصص لتحديد الثغرات - Aioga أخبار الذكاء الاصطناعي","description":"أصدرت شركة سيسكو سلسلة نماذج اللغة الصغيرة Antares، المخصصة لتحديد الثغرات، وقد تم إصدار إصدارات ذات 350 مليون و1 مليار معلمة على Hugging Face، وسيتم لاحقًا إطلاق إصدار 3 مليارات م...","url":"https://www.aioga.com/ar/news/cmrvynans02ekbipza1gsp8tb/","contentTranslated":true,"sourceHash":"ef2a104326f431fe","translatedAt":"2026-07-22T23:05:25.972Z"},"hi":{"title":"सिस्को ने Antares लघु भाषा मॉडल पेश किया, विशेष रूप से कमजोरियों का पता लगाने के लिए","summary":"सिस्को ने Antares श्रृंखला के छोटे भाषा मॉडल जारी किए हैं, जो विशेष रूप से भेद्यता पता लगाने के लिए डिज़ाइन किए गए हैं। इन्हें Hugging Face पर 350M और 1B पैरामीटर संस्करणों में जारी किया गया है, और भविष्य में 3B संस्करण लॉन्च किया जाएगा। सिस्को के भेद्यता पता लगाने के बेंचमार्क परीक्षणों में, Antares-1B ने 22.4% भेद्यता फाइल पता लगाने की दर हासिल की, जो GPT-5.5 के 22.1% से थोड़ी अधिक है। यह मॉडल स्थानीय रूप से तैनात किया जा सकता है, लेकिन केवल सहायक उद्देश्य के लिए है; कंपनियों को अभी भी अन्य सुरक्षा उपायों के साथ मिलकर पूरी सुरक्षा प्रणाली स्थापित करने की आवश्यकता है।","category":"模型更新","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"सिस्को ने Antares लघु भाषा मॉडल पेश किया, विशेष रूप से कमजोरियों का पता लगाने के लिए - Aioga AI समाचार","description":"सिस्को ने Antares श्रृंखला के छोटे भाषा मॉडल जारी किए हैं, जो विशेष रूप से भेद्यता पता लगाने के लिए डिज़ाइन किए गए हैं। इन्हें Hugging Face पर 350M और 1B पैरामीटर संस्करणों में जार...","url":"https://www.aioga.com/hi/news/cmrvynans02ekbipza1gsp8tb/","contentTranslated":true,"sourceHash":"ef2a104326f431fe","translatedAt":"2026-07-22T23:05:19.806Z"},"it":{"title":"Cisco lancia il piccolo modello linguistico Antares, specificamente progettato per il rilevamento delle vulnerabilità","summary":"Cisco ha rilasciato la serie di piccoli modelli linguistici Antares, appositamente progettata per il rilevamento delle vulnerabilità, già pubblicata su Hugging Face nelle versioni da 350M e 1B parametri, e successivamente sarà rilasciata la versione da 3B. Nei benchmark di rilevamento delle vulnerabilità di Cisco, Antares-1B ha raggiunto un tasso di rilevamento dei file vulnerabili del 22,4%, leggermente superiore al 22,1% di GPT-5.5. Questo modello può essere distribuito localmente, ma solo come supporto; le aziende devono comunque combinare altre misure di sicurezza per stabilire un sistema di protezione completo.","category":"模型更新","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Cisco lancia il piccolo modello linguistico Antares, specificamente progettato per il rilevamento delle vulnerabilità - Aioga Notizie IA","description":"Cisco ha rilasciato la serie di piccoli modelli linguistici Antares, appositamente progettata per il rilevamento delle vulnerabilità, già pubblicata su Hugging Face nelle versioni...","url":"https://www.aioga.com/it/news/cmrvynans02ekbipza1gsp8tb/","contentTranslated":true,"sourceHash":"ef2a104326f431fe","translatedAt":"2026-07-22T23:06:07.001Z"},"nl":{"title":"Cisco introduceert Antares klein taalmodel, speciaal voor het lokaliseren van kwetsbaarheden","summary":"Cisco heeft de Antares-serie van kleine taalmodellen uitgebracht, speciaal voor kwetsbaarheidsdetectie, waarbij versies van 350 miljoen en 1 miljard parameters al op Hugging Face zijn gepubliceerd, en een 3 miljard versie in de toekomst zal worden uitgebracht. In Cisco's benchmarktest voor kwetsbaarheidsdetectie bereikt Antares-1B een detectiesnelheid van 22,4% voor kwetsbare bestanden, iets hoger dan de 22,1% van GPT-5.5. Dit model kan lokaal worden ingezet, maar alleen als hulpmiddel; bedrijven moeten nog steeds andere beveiligingsmaatregelen combineren om een volledig beschermingssysteem op te bouwen.","category":"模型更新","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Cisco introduceert Antares klein taalmodel, speciaal voor het lokaliseren van kwetsbaarheden - Aioga AI-nieuws","description":"Cisco heeft de Antares-serie van kleine taalmodellen uitgebracht, speciaal voor kwetsbaarheidsdetectie, waarbij versies van 350 miljoen en 1 miljard parameters al op Hugging Face z...","url":"https://www.aioga.com/nl/news/cmrvynans02ekbipza1gsp8tb/","contentTranslated":true,"sourceHash":"ef2a104326f431fe","translatedAt":"2026-07-22T23:06:10.131Z"},"tr":{"title":"Cisco, Antares adlı küçük dil modelini tanıttı, özel olarak güvenlik açıklarının tespiti için","summary":"Cisco, özellikle güvenlik açıklarını tespit etmek için tasarlanmış Antares serisi küçük dil modellerini yayınladı; Hugging Face üzerinde 350M ve 1B parametreli versiyonları mevcuttur ve daha sonra 3B versiyonu da piyasaya sürülecek. Cisco'nun güvenlik açığı tespit kıyaslama testlerinde, Antares-1B'nin açık dosya tespit oranı %22,4 olup, GPT-5.5'in %22,1'inden biraz yüksektir. Bu model yerel olarak dağıtılabilir, ancak yalnızca yardımcı amaçlı kullanılmalıdır; şirketlerin tam bir koruma sistemi kurmak için diğer güvenlik önlemleriyle birlikte kullanmaları gerekmektedir.","category":"模型更新","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Cisco, Antares adlı küçük dil modelini tanıttı, özel olarak güvenlik açıklarının tespiti için - Aioga AI Haberleri","description":"Cisco, özellikle güvenlik açıklarını tespit etmek için tasarlanmış Antares serisi küçük dil modellerini yayınladı; Hugging Face üzerinde 350M ve 1B parametreli versiyonları mevcutt...","url":"https://www.aioga.com/tr/news/cmrvynans02ekbipza1gsp8tb/","contentTranslated":true,"sourceHash":"ef2a104326f431fe","translatedAt":"2026-07-22T23:06:52.927Z"},"vi":{"title":"Cisco ra mắt mô hình ngôn ngữ nhỏ Antares, chuyên dùng để xác định lỗ hổng","summary":"Cisco đã phát hành dòng mô hình ngôn ngữ nhỏ Antares, chuyên dùng để định vị lỗ hổng, đã ra mắt phiên bản 350M và 1B trên Hugging Face, và phiên bản 3B sẽ được ra mắt sau. Trong các bài kiểm tra đánh giá định vị lỗ hổng của Cisco, tỉ lệ phát hiện tệp lỗ hổng của Antares-1B đạt 22,4%, cao hơn một chút so với 22,1% của GPT-5.5. Mô hình này có thể triển khai tại chỗ, nhưng chỉ dùng để hỗ trợ, doanh nghiệp vẫn cần kết hợp các biện pháp bảo mật khác để xây dựng hệ thống phòng thủ hoàn chỉnh.","category":"模型更新","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Cisco ra mắt mô hình ngôn ngữ nhỏ Antares, chuyên dùng để xác định lỗ hổng - Tin tức AI Aioga","description":"Cisco đã phát hành dòng mô hình ngôn ngữ nhỏ Antares, chuyên dùng để định vị lỗ hổng, đã ra mắt phiên bản 350M và 1B trên Hugging Face, và phiên bản 3B sẽ được ra mắt sau. Trong cá...","url":"https://www.aioga.com/vi/news/cmrvynans02ekbipza1gsp8tb/","contentTranslated":true,"sourceHash":"ef2a104326f431fe","translatedAt":"2026-07-22T23:06:49.716Z"},"id":{"title":"Cisco meluncurkan model bahasa kecil Antares, khusus untuk penentuan lokasi kerentanan","summary":"Cisco merilis seri model bahasa kecil Antares, khusus digunakan untuk penentuan lokasi kerentanan, telah merilis versi 350M dan 1B parameter di Hugging Face, versi 3B akan dirilis kemudian. Dalam uji benchmark penentuan lokasi kerentanan Cisco, tingkat deteksi file kerentanan Antares-1B mencapai 22,4%, sedikit lebih tinggi dari 22,1% milik GPT-5.5. Model ini dapat diterapkan secara lokal, tetapi hanya untuk tujuan bantuan, perusahaan tetap perlu menggabungkan langkah-langkah keamanan lain untuk membangun sistem pertahanan yang lengkap.","category":"模型更新","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Cisco meluncurkan model bahasa kecil Antares, khusus untuk penentuan lokasi kerentanan - Berita AI Aioga","description":"Cisco merilis seri model bahasa kecil Antares, khusus digunakan untuk penentuan lokasi kerentanan, telah merilis versi 350M dan 1B parameter di Hugging Face, versi 3B akan dirilis...","url":"https://www.aioga.com/id/news/cmrvynans02ekbipza1gsp8tb/","contentTranslated":true,"sourceHash":"ef2a104326f431fe","translatedAt":"2026-07-22T23:07:30.107Z"},"th":{"title":"Cisco เปิดตัวโมเดลภาษาขนาดเล็ก Antares ใช้สำหรับระบุจุดบกพร่อง","summary":"ซิสโก้เปิดตัวโมเดลภาษาขนาดเล็กซีรีส์ Antares ซึ่งออกแบบมาสำหรับการระบุช่องโหว่โดยเฉพาะ โดยได้เผยแพร่รุ่นที่มีพารามิเตอร์ 350M และ 1B บน Hugging Face และในอนาคตรุ่น 3B จะตามมา ในการทดสอบมาตรฐานการระบุช่องโหว่ของซิสโก้ Antares-1B มีอัตราการตรวจจับไฟล์ช่องโหว่ถึง 22.4% ซึ่งสูงกว่า GPT-5.5 ที่ 22.1% เล็กน้อย โมเดลดังกล่าวสามารถติดตั้งใช้งานในเครื่องได้ แต่ใช้เพื่อช่วยสนับสนุนเท่านั้น บริษัทต่าง ๆ ยังคงต้องรวมมาตรการรักษาความปลอดภัยอื่น ๆ เพื่อสร้างระบบป้องกันที่ครบถ้วน","category":"模型更新","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Cisco เปิดตัวโมเดลภาษาขนาดเล็ก Antares ใช้สำหรับระบุจุดบกพร่อง - ข่าว AI Aioga","description":"ซิสโก้เปิดตัวโมเดลภาษาขนาดเล็กซีรีส์ Antares ซึ่งออกแบบมาสำหรับการระบุช่องโหว่โดยเฉพาะ โดยได้เผยแพร่รุ่นที่มีพารามิเตอร์ 350M และ 1B บน Hugging Face และในอนาคตรุ่น 3B จะตามมา ในการ...","url":"https://www.aioga.com/th/news/cmrvynans02ekbipza1gsp8tb/","contentTranslated":true,"sourceHash":"ef2a104326f431fe","translatedAt":"2026-07-22T23:07:37.699Z"},"pl":{"title":"Cisco wprowadza mały model językowy Antares, przeznaczony specjalnie do lokalizowania luk","summary":"Cisco wydało serię małych modeli językowych Antares, specjalnie przeznaczonych do lokalizowania luk, dostępne na Hugging Face w wersjach 350M i 1B parametrów, a w przyszłości zostanie wydana wersja 3B. W benchmarku lokalizacji luk Cisco, Antares-1B osiągnął 22,4% wykrywalności plików z lukami, nieco wyższą niż 22,1% dla GPT-5.5. Model można wdrożyć lokalnie, ale służy wyłącznie jako wsparcie; przedsiębiorstwa wciąż muszą stosować inne środki bezpieczeństwa, aby stworzyć pełny system ochrony.","category":"模型更新","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Cisco wprowadza mały model językowy Antares, przeznaczony specjalnie do lokalizowania luk - Aioga Wiadomości AI","description":"Cisco wydało serię małych modeli językowych Antares, specjalnie przeznaczonych do lokalizowania luk, dostępne na Hugging Face w wersjach 350M i 1B parametrów, a w przyszłości zosta...","url":"https://www.aioga.com/pl/news/cmrvynans02ekbipza1gsp8tb/","contentTranslated":true,"sourceHash":"ef2a104326f431fe","translatedAt":"2026-07-22T23:08:17.320Z"}}}}