{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-08-11T09:21:12.743Z","headline":"Mistral 推出 Shieldstral 多模态内容审核模型，单张 16GB GPU 可运行","description":"Mistral AI 发布 Shieldstral 内容审核模型，总参数量 3B，采用开放权重并依据 Apache 2.0 许可证发布，已上线 Hugging Face，支持 12 种语言，可在单张 16GB GPU 上运行。","url":"https://www.aioga.com/news/cmsfkhnwx035xrochw79mida5/","mainEntityOfPage":"https://www.aioga.com/news/cmsfkhnwx035xrochw79mida5/","datePublished":"2026-08-05T03:58:03.000Z","dateModified":"2026-08-05T03:58:03.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://www.ithome.com/0/985/886.htm","https://aihot.virxact.com/items/cmsfkhnwx035xrochw79mida5"],"canonicalUrl":"https://www.aioga.com/news/cmsfkhnwx035xrochw79mida5/","directAnswer":{"@type":"Answer","text":"Mistral AI 发布 Shieldstral 多模态内容审核模型，总参数量为 3B，采用开放权重并依据 Apache 2.0 许可证发布。该模型已上线 Hugging Face，支持 12 种语言，并可在单张 16GB GPU 上运行。","url":"https://www.aioga.com/news/cmsfkhnwx035xrochw79mida5/","dateCreated":"2026-08-05T03:58:03.000Z","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/0/985/886.htm","datePublished":"2026-08-05T03:58:03.000Z","provider":{"@type":"Organization","name":"IT之家","url":"https://www.ithome.com/0/985/886.htm"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.virxact.com/items/cmsfkhnwx035xrochw79mida5","datePublished":"2026-08-05T03:58:03.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.virxact.com/items/cmsfkhnwx035xrochw79mida5"}}],"aggregationSource":"IT之家（RSS）","originalPublisher":{"name":"IT之家","url":"https://www.ithome.com/0/985/886.htm"},"geoDeepAnswer":null,"article":{"id":"cmsfkhnwx035xrochw79mida5","slug":"cmsfkhnwx035xrochw79mida5","url":"https://www.aioga.com/news/cmsfkhnwx035xrochw79mida5/","title":"Mistral 推出 Shieldstral 多模态内容审核模型，单张 16GB GPU 可运行","title_en":"最强多模态内容审核开源 AI 模型：Mistral 推出 Shieldstral，单张 16GB GPU 可运行","summary":"Mistral AI 发布 Shieldstral 内容审核模型，总参数量 3B，采用开放权重并依据 Apache 2.0 许可证发布，已上线 Hugging Face，支持 12 种语言，可在单张 16GB GPU 上运行。","source":"IT之家（RSS）","sourceUrl":"https://www.ithome.com/0/985/886.htm","aiHotUrl":"https://aihot.virxact.com/items/cmsfkhnwx035xrochw79mida5","publishedAt":"2026-08-05T03:58:03.000Z","category":"模型更新","score":64,"selected":false,"articleBody":["IT之家：https://www.ithome.com/ 8 月 5 日消息，Mistral AI 昨日（8 月 4 日）发布公告，宣布推出 Shieldstral 内容审核 AI 模型，总参数量为 3B（30 亿），采用开放权重，依据 Apache 2.0 许可证发布。","模型已上线 Hugging Face 平台，支持 12 种语言，可在单张 16GB GPU 上运行。Mistral 表示该模型在内容安全方面，性能媲美 7 倍规模的开放模型， 并且在多模态内容审核领域实现 SOTA。","IT之家注：SOTA 全称为 state-of-the-art，是指在特定人工智能任务或基准测试中表现最优秀、水平最先进的模型或算法集合，它会随着技术迭代和新研究的发布而不断变化。","Mistral 指出，多数防护模型会把伤害类别体系固化在模型权重中。产品更换使用场景后，开发者通常需要重新训练模型。","Shieldstral 则把审核政策放入输入内容：操作者可以写入一个“是或否”问题，再提供评估场景和严格程度说明，模型随后从单个输出词元中生成经过校准的安全分数。","：提供待审核内容，可以是提示词、回答、提示词与回答组合，也可以是附带可选文本的图像。","推理时，模型只读取“是”和“否”两个词元的逻辑值，再通过 softmax 归一化为连续分数，并以 0.5 为阈值输出二元判断。该机制可覆盖提示词分类、回答审核、拒答检测和毒性检测。"],"articleImages":[{"sourceUrl":"https://img.ithome.com/newsuploadfiles/2026/8/902eaf3f-8b33-4f5a-b20e-fc1a51ad2f65.png?x-bce-process=image/format,f_auto","alt":"","afterParagraph":1,"url":"/media/articles/cmsfkhnwx035xrochw79mida5/607ec4b165d8fbc1.webp"},{"sourceUrl":"https://img.ithome.com/newsuploadfiles/2026/8/ec15068f-1c8d-4449-8f4a-b323a9dd0bf7.png?x-bce-process=image/format,f_auto","alt":"","afterParagraph":2,"url":"/media/articles/cmsfkhnwx035xrochw79mida5/feca2034ad692e2d.png"},{"sourceUrl":"https://img.ithome.com/newsuploadfiles/2026/8/af663e98-d92a-4dd1-b2b7-3c7437f351bd.png?x-bce-process=image/format,f_auto","alt":"","afterParagraph":6,"url":"/media/articles/cmsfkhnwx035xrochw79mida5/4c862113309fcd8d.png"}],"mediaStatus":"ok","articleBodyZh":["Mistral 推出 Shieldstral 多模态内容审核模型，单张 16GB GPU 可运行 这条更新来自 ithome.com，发布时间为 2026-08-05，Aioga 保留原文入口以便核验。","摘要：Mistral AI 发布 Shieldstral 内容审核模型，总参数量 3B，采用开放权重并依据 Apache 2.0 许可证发布，已上线 Hugging Face，支持 12 种语言，可在单张 16GB GPU 上运行。","背景：Shieldstral 将审核政策放入输入内容，操作者可设定是或否问题、评估场景与严格程度。模型针对提示词、回答、提示词与回答组合及带可选文本的图像进行判断，并通过“是”“否”逻辑值与 softmax 生成安全分数。","Aioga 观察：Aioga 判断，将审核政策置于输入内容，意味着模型可通过调整问题、场景和严格程度适配不同审核要求。Mistral 称其性能可媲美规模大七倍的开放模型，并在多模态审核领域实现 SOTA，但材料未提供具体基准与测试数据。","影响与后续：开放权重、Apache 2.0 许可证及单张 16GB GPU 的运行条件，可能降低部分开发者部署审核能力的门槛。支持 12 种语言和可配置审核政策，也值得关注其在不同语言、场景及严格程度下的实际一致性。 Aioga 建议后续关注 Hugging Face 页面及 Mistral 公告中的评测细节，重点核对多模态任务、12 种语言、不同审核政策和 0.5 阈值下的表现，再判断其是否适合具体业务流程。"],"translationStatus":"","bodyOrigin":"source-page","editorial":{"summary":"Mistral AI 发布 Shieldstral 多模态内容审核模型，总参数量为 3B，采用开放权重并依据 Apache 2.0 许可证发布。该模型已上线 Hugging Face，支持 12 种语言，并可在单张 16GB GPU 上运行。","background":"Shieldstral 将审核政策放入输入内容，操作者可设定是或否问题、评估场景与严格程度。模型针对提示词、回答、提示词与回答组合及带可选文本的图像进行判断，并通过“是”“否”逻辑值与 softmax 生成安全分数。","viewpoint":"Aioga 判断，将审核政策置于输入内容，意味着模型可通过调整问题、场景和严格程度适配不同审核要求。Mistral 称其性能可媲美规模大七倍的开放模型，并在多模态审核领域实现 SOTA，但材料未提供具体基准与测试数据。","implications":"开放权重、Apache 2.0 许可证及单张 16GB GPU 的运行条件，可能降低部分开发者部署审核能力的门槛。支持 12 种语言和可配置审核政策，也值得关注其在不同语言、场景及严格程度下的实际一致性。","nextStep":"Aioga 建议后续关注 Hugging Face 页面及 Mistral 公告中的评测细节，重点核对多模态任务、12 种语言、不同审核政策和 0.5 阈值下的表现，再判断其是否适合具体业务流程。","evidenceRefs":["title","summary","articleBody","source"],"status":"published","aiGenerated":true,"autoApproved":true,"generatedBy":"aioga-editorial:gpt-5.6-sol","reviewedBy":"aioga-editorial-review:gpt-5.6-sol","generatedAt":"2026-08-05T04:31:03.187Z","sourceHash":"efbd0a8c1acdef16","review":{"approved":true,"groundedness":97,"clarity":93,"duplicationRisk":24,"blockingIssues":[],"notes":["候选内容准确区分了 Mistral 的性能主张与编辑判断，并明确指出来源材料未提供具体基准和测试数据。","如需进一步完整，可在 background 中补充模型以 0.5 为默认阈值输出二元判断，但这不影响当前内容通过审核。","“可能降低部分开发者部署审核能力的门槛”属于合理且已作不确定性限定的推论，不构成事实性断言。"]},"validation":{"passed":true,"mode":"ai-auto","revisions":0,"checks":["schema","length","source-attribution","low-source-overlap","no-html","independent-ai-review"]}},"tags":["模型更新","IT之家（RSS）"],"translations":{"zh-CN":{"title":"Mistral 推出 Shieldstral 多模态内容审核模型，单张 16GB GPU 可运行","summary":"Mistral AI 发布 Shieldstral 内容审核模型，总参数量 3B，采用开放权重并依据 Apache 2.0 许可证发布，已上线 Hugging Face，支持 12 种语言，可在单张 16GB GPU 上运行。","category":"模型更新","source":"IT之家","aggregationSource":"IT之家（RSS）","pageTitle":"Mistral 推出 Shieldstral 多模态内容审核模型，单张 16GB GPU 可运行 - Aioga AI资讯","description":"Mistral AI 发布 Shieldstral 内容审核模型，总参数量 3B，采用开放权重并依据 Apache 2.0 许可证发布，已上线 Hugging Face，支持 12 种语言，可在单张 16GB GPU 上运行。","url":"https://www.aioga.com/news/cmsfkhnwx035xrochw79mida5/","articleBody":["Mistral 推出 Shieldstral 多模态内容审核模型，单张 16GB GPU 可运行 这条更新来自 ithome.com，发布时间为 2026-08-05，Aioga 保留原文入口以便核验。","摘要：Mistral AI 发布 Shieldstral 内容审核模型，总参数量 3B，采用开放权重并依据 Apache 2.0 许可证发布，已上线 Hugging Face，支持 12 种语言，可在单张 16GB GPU 上运行。","背景：Shieldstral 将审核政策放入输入内容，操作者可设定是或否问题、评估场景与严格程度。模型针对提示词、回答、提示词与回答组合及带可选文本的图像进行判断，并通过“是”“否”逻辑值与 softmax 生成安全分数。","Aioga 观察：Aioga 判断，将审核政策置于输入内容，意味着模型可通过调整问题、场景和严格程度适配不同审核要求。Mistral 称其性能可媲美规模大七倍的开放模型，并在多模态审核领域实现 SOTA，但材料未提供具体基准与测试数据。","影响与后续：开放权重、Apache 2.0 许可证及单张 16GB GPU 的运行条件，可能降低部分开发者部署审核能力的门槛。支持 12 种语言和可配置审核政策，也值得关注其在不同语言、场景及严格程度下的实际一致性。 Aioga 建议后续关注 Hugging Face 页面及 Mistral 公告中的评测细节，重点核对多模态任务、12 种语言、不同审核政策和 0.5 阈值下的表现，再判断其是否适合具体业务流程。"]},"en":{"title":"Mistral launches the Shieldstral multimodal content moderation model, which can run on a single 16GB GPU","summary":"Mistral AI has released the Shieldstral content moderation model, with a total of 3 billion parameters, using open weights and released under the Apache 2.0 license. It is now available on Hugging Face, supports 12 languages, and can run on a single 16GB GPU.","category":"Models","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Mistral launches the Shieldstral multimodal content moderation model, which can run on a single 16GB GPU - Aioga AI News","description":"Mistral AI has released the Shieldstral content moderation model, with a total of 3 billion parameters, using open weights and released under the Apache 2.0 license. It is now avai...","url":"https://www.aioga.com/en/news/cmsfkhnwx035xrochw79mida5/","contentTranslated":true,"sourceHash":"85cebad4fd55313b","translatedAt":"2026-08-05T04:22:09.941Z"},"ja":{"title":"MistralはShieldstral多モーダルコンテンツ審査モデルを発表、単一16GB GPUで動作可能","summary":"Mistral AIはShieldstralコンテンツ監査モデルを公開しました。総パラメータ数は3Bで、オープンウェイトを採用し、Apache 2.0ライセンスに基づいて公開されています。Hugging Faceで利用可能で、12言語に対応しており、単一の16GB GPUで実行可能です。","category":"モデル更新","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"MistralはShieldstral多モーダルコンテンツ審査モデルを発表、単一16GB GPUで動作可能 - Aioga AIニュース","description":"Mistral AIはShieldstralコンテンツ監査モデルを公開しました。総パラメータ数は3Bで、オープンウェイトを採用し、Apache 2.0ライセンスに基づいて公開されています。Hugging Faceで利用可能で、12言語に対応しており、単一の16GB GPUで実行可能です。","url":"https://www.aioga.com/ja/news/cmsfkhnwx035xrochw79mida5/","contentTranslated":true,"sourceHash":"85cebad4fd55313b","translatedAt":"2026-08-05T04:22:29.791Z"},"ko":{"title":"Mistral이 Shieldstral 다중 모드 콘텐츠 검열 모델을 출시했으며 단일 16GB GPU로 실행할 수 있습니다","summary":"Mistral AI는 Shieldstral 콘텐츠 검열 모델을 출시했으며, 전체 매개변수 수는 3B입니다. 오픈 가중치를 사용하고 Apache 2.0 라이선스에 따라 공개되었으며, Hugging Face에서 이미 사용 가능하며 12가지 언어를 지원하고 단일 16GB GPU에서 실행할 수 있습니다.","category":"모델 업데이트","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Mistral이 Shieldstral 다중 모드 콘텐츠 검열 모델을 출시했으며 단일 16GB GPU로 실행할 수 있습니다 - Aioga AI 뉴스","description":"Mistral AI는 Shieldstral 콘텐츠 검열 모델을 출시했으며, 전체 매개변수 수는 3B입니다. 오픈 가중치를 사용하고 Apache 2.0 라이선스에 따라 공개되었으며, Hugging Face에서 이미 사용 가능하며 12가지 언어를 지원하고 단일 16GB GPU에서 실행할 수 있습니다.","url":"https://www.aioga.com/ko/news/cmsfkhnwx035xrochw79mida5/","contentTranslated":true,"sourceHash":"85cebad4fd55313b","translatedAt":"2026-08-05T04:23:24.112Z"},"es":{"title":"Mistral lanza el modelo de revisión de contenido multimodal Shieldstral, que puede ejecutarse en una sola GPU de 16GB","summary":"Mistral AI lanzó el modelo de revisión de contenido Shieldstral, con un total de 3 mil millones de parámetros, utilizando pesos abiertos y publicado bajo la licencia Apache 2.0, ya disponible en Hugging Face, soporta 12 idiomas y puede ejecutarse en una sola GPU de 16GB.","category":"Modelos","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Mistral lanza el modelo de revisión de contenido multimodal Shieldstral, que puede ejecutarse en una sola GPU de 16GB - Aioga Noticias de IA","description":"Mistral AI lanzó el modelo de revisión de contenido Shieldstral, con un total de 3 mil millones de parámetros, utilizando pesos abiertos y publicado bajo la licencia Apache 2.0, ya...","url":"https://www.aioga.com/es/news/cmsfkhnwx035xrochw79mida5/","contentTranslated":true,"sourceHash":"85cebad4fd55313b","translatedAt":"2026-08-05T04:23:18.914Z"},"fr":{"title":"Mistral lance le modèle de modération de contenu multimodal Shieldstral, pouvant fonctionner sur un seul GPU de 16 Go","summary":"Mistral AI a publié le modèle de modération de contenu Shieldstral, avec un total de 3 milliards de paramètres, utilisant des poids ouverts et publié sous licence Apache 2.0, déjà disponible sur Hugging Face, prenant en charge 12 langues et pouvant fonctionner sur un seul GPU de 16 Go.","category":"Modèles","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Mistral lance le modèle de modération de contenu multimodal Shieldstral, pouvant fonctionner sur un seul GPU de 16 Go - Aioga Actualités IA","description":"Mistral AI a publié le modèle de modération de contenu Shieldstral, avec un total de 3 milliards de paramètres, utilisant des poids ouverts et publié sous licence Apache 2.0, déjà...","url":"https://www.aioga.com/fr/news/cmsfkhnwx035xrochw79mida5/","contentTranslated":true,"sourceHash":"85cebad4fd55313b","translatedAt":"2026-08-05T04:24:11.185Z"},"de":{"title":"Mistral bringt das multimodale Inhaltsprüfmodell Shieldstral auf den Markt, das auf einer einzigen 16-GB-GPU betrieben werden kann","summary":"Mistral AI veröffentlicht das Inhaltsprüfungsmodell Shieldstral, mit insgesamt 3 Milliarden Parametern, es verwendet offene Gewichte und wird unter der Apache 2.0 Lizenz veröffentlicht, ist auf Hugging Face verfügbar, unterstützt 12 Sprachen und kann auf einer einzelnen 16GB GPU betrieben werden.","category":"模型更新","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Mistral bringt das multimodale Inhaltsprüfmodell Shieldstral auf den Markt, das auf einer einzigen 16-GB-GPU betrieben werden kann - Aioga KI-News","description":"Mistral AI veröffentlicht das Inhaltsprüfungsmodell Shieldstral, mit insgesamt 3 Milliarden Parametern, es verwendet offene Gewichte und wird unter der Apache 2.0 Lizenz veröffentl...","url":"https://www.aioga.com/de/news/cmsfkhnwx035xrochw79mida5/","contentTranslated":true,"sourceHash":"85cebad4fd55313b","translatedAt":"2026-08-05T04:24:08.826Z"},"pt-BR":{"title":"Mistral lança o modelo de moderação de conteúdo multimodal Shieldstral, podendo ser executado em uma única GPU de 16GB","summary":"Mistral AI lançou o modelo de moderação de conteúdo Shieldstral, com um total de 3 bilhões de parâmetros, disponibilizado com pesos abertos e sob a licença Apache 2.0, já disponível no Hugging Face, suportando 12 idiomas e podendo ser executado em uma única GPU de 16GB.","category":"模型更新","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Mistral lança o modelo de moderação de conteúdo multimodal Shieldstral, podendo ser executado em uma única GPU de 16GB - Aioga Notícias de IA","description":"Mistral AI lançou o modelo de moderação de conteúdo Shieldstral, com um total de 3 bilhões de parâmetros, disponibilizado com pesos abertos e sob a licença Apache 2.0, já disponíve...","url":"https://www.aioga.com/pt-BR/news/cmsfkhnwx035xrochw79mida5/","contentTranslated":true,"sourceHash":"85cebad4fd55313b","translatedAt":"2026-08-05T04:24:55.492Z"},"ru":{"title":"Mistral выпустила мульти-модальную модель модерации контента Shieldstral, которая может работать на одной 16GB GPU","summary":"Mistral AI выпустила модель модерации контента Shieldstral, общий объем параметров 3B, использует открытые веса и публикуется в соответствии с лицензией Apache 2.0, уже доступна на Hugging Face, поддерживает 12 языков, может работать на одной 16GB GPU.","category":"模型更新","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Mistral выпустила мульти-модальную модель модерации контента Shieldstral, которая может работать на одной 16GB GPU - Aioga Новости ИИ","description":"Mistral AI выпустила модель модерации контента Shieldstral, общий объем параметров 3B, использует открытые веса и публикуется в соответствии с лицензией Apache 2.0, уже доступна на...","url":"https://www.aioga.com/ru/news/cmsfkhnwx035xrochw79mida5/","contentTranslated":true,"sourceHash":"85cebad4fd55313b","translatedAt":"2026-08-05T04:25:00.167Z"},"ar":{"title":"أطلقت Mistral نموذج مراجعة المحتوى متعدد الوسائط Shieldstral، يمكن تشغيله على بطاقة GPU واحدة بسعة 16 جيجابايت","summary":"أصدرت Mistral AI نموذج مراجعة المحتوى Shieldstral، بإجمالي 3 مليارات معلمة، مع وزن مفتوح ومرخص بموجب رخصة Apache 2.0، متاح على Hugging Face، ويدعم 12 لغة، ويمكن تشغيله على وحدة GPU واحدة بسعة 16 جيجابايت.","category":"模型更新","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"أطلقت Mistral نموذج مراجعة المحتوى متعدد الوسائط Shieldstral، يمكن تشغيله على بطاقة GPU واحدة بسعة 16 جيجابايت - Aioga أخبار الذكاء الاصطناعي","description":"أصدرت Mistral AI نموذج مراجعة المحتوى Shieldstral، بإجمالي 3 مليارات معلمة، مع وزن مفتوح ومرخص بموجب رخصة Apache 2.0، متاح على Hugging Face، ويدعم 12 لغة، ويمكن تشغيله على وحدة GPU...","url":"https://www.aioga.com/ar/news/cmsfkhnwx035xrochw79mida5/","contentTranslated":true,"sourceHash":"85cebad4fd55313b","translatedAt":"2026-08-05T04:25:50.577Z"},"hi":{"title":"Mistral ने Shieldstral बहुमोडल सामग्री समीक्षा मॉडल लॉन्च किया, एकल 16GB GPU पर चल सकता है","summary":"Mistral AI ने Shieldstral सामग्री समीक्षा मॉडल जारी किया, कुल पैरामीटर 3B है, खुला वज़न अपनाया गया और Apache 2.0 लाइसेंस के तहत जारी किया गया, Hugging Face पर उपलब्ध है, 12 भाषाओं का समर्थन करता है, और एक 16GB GPU पर चल सकता है।","category":"模型更新","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Mistral ने Shieldstral बहुमोडल सामग्री समीक्षा मॉडल लॉन्च किया, एकल 16GB GPU पर चल सकता है - Aioga AI समाचार","description":"Mistral AI ने Shieldstral सामग्री समीक्षा मॉडल जारी किया, कुल पैरामीटर 3B है, खुला वज़न अपनाया गया और Apache 2.0 लाइसेंस के तहत जारी किया गया, Hugging Face पर उपलब्ध है, 12 भाषाओं...","url":"https://www.aioga.com/hi/news/cmsfkhnwx035xrochw79mida5/","contentTranslated":true,"sourceHash":"85cebad4fd55313b","translatedAt":"2026-08-05T04:25:52.502Z"},"it":{"title":"Mistral lancia il modello di revisione dei contenuti multimodale Shieldstral, operabile su una singola GPU da 16GB","summary":"Mistral AI ha rilasciato il modello di revisione dei contenuti Shieldstral, con un totale di 3 miliardi di parametri, utilizzando pesi aperti e rilasciato secondo la licenza Apache 2.0, già disponibile su Hugging Face, supporta 12 lingue e può essere eseguito su una singola GPU da 16 GB.","category":"模型更新","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Mistral lancia il modello di revisione dei contenuti multimodale Shieldstral, operabile su una singola GPU da 16GB - Aioga Notizie IA","description":"Mistral AI ha rilasciato il modello di revisione dei contenuti Shieldstral, con un totale di 3 miliardi di parametri, utilizzando pesi aperti e rilasciato secondo la licenza Apache...","url":"https://www.aioga.com/it/news/cmsfkhnwx035xrochw79mida5/","contentTranslated":true,"sourceHash":"85cebad4fd55313b","translatedAt":"2026-08-05T04:26:41.826Z"},"nl":{"title":"Mistral introduceert het multimodale inhoudsbeoordelingsmodel Shieldstral, dat kan draaien op een enkele 16GB GPU","summary":"Mistral AI heeft het contentmoderatiemodel Shieldstral uitgebracht, met een totaal aantal parameters van 3B, gebruikmakend van open weights en uitgebracht onder de Apache 2.0-licentie, beschikbaar op Hugging Face, ondersteunt 12 talen en kan draaien op een enkele 16GB GPU.","category":"模型更新","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Mistral introduceert het multimodale inhoudsbeoordelingsmodel Shieldstral, dat kan draaien op een enkele 16GB GPU - Aioga AI-nieuws","description":"Mistral AI heeft het contentmoderatiemodel Shieldstral uitgebracht, met een totaal aantal parameters van 3B, gebruikmakend van open weights en uitgebracht onder de Apache 2.0-licen...","url":"https://www.aioga.com/nl/news/cmsfkhnwx035xrochw79mida5/","contentTranslated":true,"sourceHash":"85cebad4fd55313b","translatedAt":"2026-08-05T04:26:33.875Z"},"tr":{"title":"Mistral, Shieldstral çok modlu içerik denetim modelini piyasaya sürdü, tek bir 16GB GPU ile çalıştırılabilir","summary":"Mistral AI, toplam parametre sayısı 3B olan Shieldstral içerik denetim modelini yayınladı, açık ağırlıklar kullanıyor ve Apache 2.0 lisansına göre yayınlandı, Hugging Face üzerinde yayında, 12 dili destekliyor ve tek bir 16GB GPU üzerinde çalışabiliyor.","category":"模型更新","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Mistral, Shieldstral çok modlu içerik denetim modelini piyasaya sürdü, tek bir 16GB GPU ile çalıştırılabilir - Aioga AI Haberleri","description":"Mistral AI, toplam parametre sayısı 3B olan Shieldstral içerik denetim modelini yayınladı, açık ağırlıklar kullanıyor ve Apache 2.0 lisansına göre yayınlandı, Hugging Face üzerinde...","url":"https://www.aioga.com/tr/news/cmsfkhnwx035xrochw79mida5/","contentTranslated":true,"sourceHash":"85cebad4fd55313b","translatedAt":"2026-08-05T04:27:31.229Z"},"vi":{"title":"Mistral ra mắt mô hình kiểm duyệt nội dung đa phương thức Shieldstral, có thể chạy trên GPU 16GB đơn","summary":"Mistral AI phát hành mô hình kiểm duyệt nội dung Shieldstral, tổng số tham số 3B, sử dụng trọng số mở và được phát hành theo giấy phép Apache 2.0, đã có trên Hugging Face, hỗ trợ 12 ngôn ngữ, có thể chạy trên một GPU 16GB.","category":"模型更新","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Mistral ra mắt mô hình kiểm duyệt nội dung đa phương thức Shieldstral, có thể chạy trên GPU 16GB đơn - Tin tức AI Aioga","description":"Mistral AI phát hành mô hình kiểm duyệt nội dung Shieldstral, tổng số tham số 3B, sử dụng trọng số mở và được phát hành theo giấy phép Apache 2.0, đã có trên Hugging Face, hỗ trợ 1...","url":"https://www.aioga.com/vi/news/cmsfkhnwx035xrochw79mida5/","contentTranslated":true,"sourceHash":"85cebad4fd55313b","translatedAt":"2026-08-05T04:27:24.147Z"},"id":{"title":"Mistral meluncurkan model peninjauan konten multimodal Shieldstral, dapat dijalankan dengan satu GPU 16GB","summary":"Mistral AI merilis model moderasi konten Shieldstral, dengan total 3 miliar parameter, menggunakan bobot terbuka dan diterbitkan berdasarkan lisensi Apache 2.0, sudah tersedia di Hugging Face, mendukung 12 bahasa, dan dapat dijalankan pada satu GPU 16GB.","category":"模型更新","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Mistral meluncurkan model peninjauan konten multimodal Shieldstral, dapat dijalankan dengan satu GPU 16GB - Berita AI Aioga","description":"Mistral AI merilis model moderasi konten Shieldstral, dengan total 3 miliar parameter, menggunakan bobot terbuka dan diterbitkan berdasarkan lisensi Apache 2.0, sudah tersedia di H...","url":"https://www.aioga.com/id/news/cmsfkhnwx035xrochw79mida5/","contentTranslated":true,"sourceHash":"85cebad4fd55313b","translatedAt":"2026-08-05T04:28:14.401Z"},"th":{"title":"Mistral เปิดตัวโมเดลตรวจสอบเนื้อหาแบบหลายโหมด Shieldstral สามารถรันบน GPU ขนาด 16GB ต่อหนึ่งการ์ด","summary":"Mistral AI เปิดตัวโมเดลตรวจสอบเนื้อหา Shieldstral โดยมีจำนวนพารามิเตอร์ทั้งหมด 3 พันล้าน ใช้น้ำหนักแบบเปิดและเผยแพร่ตามใบอนุญาต Apache 2.0 พร้อมใช้งานบน Hugging Face รองรับ 12 ภาษา และสามารถรันบน GPU ขนาด 16GB เพียงตัวเดียวได้","category":"模型更新","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Mistral เปิดตัวโมเดลตรวจสอบเนื้อหาแบบหลายโหมด Shieldstral สามารถรันบน GPU ขนาด 16GB ต่อหนึ่งการ์ด - ข่าว AI Aioga","description":"Mistral AI เปิดตัวโมเดลตรวจสอบเนื้อหา Shieldstral โดยมีจำนวนพารามิเตอร์ทั้งหมด 3 พันล้าน ใช้น้ำหนักแบบเปิดและเผยแพร่ตามใบอนุญาต Apache 2.0 พร้อมใช้งานบน Hugging Face รองรับ 12 ภาษา...","url":"https://www.aioga.com/th/news/cmsfkhnwx035xrochw79mida5/","contentTranslated":true,"sourceHash":"85cebad4fd55313b","translatedAt":"2026-08-05T04:28:25.422Z"},"pl":{"title":"Mistral wprowadza model kontroli treści wielomodalnej Shieldstral, który może działać na pojedynczym 16GB GPU","summary":"Mistral AI wydało model do moderacji treści Shieldstral, o łącznej liczbie parametrów 3 miliardy, z otwartymi wagami i udostępniony na licencji Apache 2.0, już dostępny na Hugging Face, obsługujący 12 języków, może działać na jednej karcie GPU 16GB.","category":"模型更新","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Mistral wprowadza model kontroli treści wielomodalnej Shieldstral, który może działać na pojedynczym 16GB GPU - Aioga Wiadomości AI","description":"Mistral AI wydało model do moderacji treści Shieldstral, o łącznej liczbie parametrów 3 miliardy, z otwartymi wagami i udostępniony na licencji Apache 2.0, już dostępny na Hugging...","url":"https://www.aioga.com/pl/news/cmsfkhnwx035xrochw79mida5/","contentTranslated":true,"sourceHash":"85cebad4fd55313b","translatedAt":"2026-08-05T04:29:18.425Z"}}}}