{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-08-31T15:40:19.115Z","headline":"DeepSeek-V4-Flash-Vision-Exp 模型已开源，多模态 Agent 能力接近 Opus-4.8","description":"DeepSeek 于 8 月 31 日在 Hugging Face 开源首个多模态模型 DeepSeek-V4-Flash-Vision-Exp，采用 MIT License，公开模型文件、Tokenizer、Prompt Encoding 参考实现及最小化 PyTorch 推理实现。","url":"https://www.aioga.com/news/cmth7tmq2067orodmh6g0sxie/","mainEntityOfPage":"https://www.aioga.com/news/cmth7tmq2067orodmh6g0sxie/","datePublished":"2026-08-31T11:35:24.000Z","dateModified":"2026-08-31T11:35:24.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://www.ithome.com/0/996/637.htm","https://aihot.virxact.com/items/cmth7tmq2067orodmh6g0sxie"],"canonicalUrl":"https://www.aioga.com/news/cmth7tmq2067orodmh6g0sxie/","directAnswer":{"@type":"Answer","text":"Hugging Face 已公开 DeepSeek V4 系列的首个视觉模型实验版本，发布内容覆盖权重、分词器、提示编码参考代码和基础推理实现，并采用 MIT License。","url":"https://www.aioga.com/news/cmth7tmq2067orodmh6g0sxie/","dateCreated":"2026-08-31T11:35:24.000Z","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/0/996/637.htm","datePublished":"2026-08-31T11:35:24.000Z","provider":{"@type":"Organization","name":"IT之家","url":"https://www.ithome.com/0/996/637.htm"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.virxact.com/items/cmth7tmq2067orodmh6g0sxie","datePublished":"2026-08-31T11:35:24.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.virxact.com/items/cmth7tmq2067orodmh6g0sxie"}}],"aggregationSource":"IT之家（RSS）","originalPublisher":{"name":"IT之家","url":"https://www.ithome.com/0/996/637.htm"},"geoDeepAnswer":null,"article":{"id":"cmth7tmq2067orodmh6g0sxie","slug":"cmth7tmq2067orodmh6g0sxie","url":"https://www.aioga.com/news/cmth7tmq2067orodmh6g0sxie/","title":"DeepSeek-V4-Flash-Vision-Exp 模型已开源，多模态 Agent 能力接近 Opus-4.8","title_en":"","summary":"DeepSeek 于 8 月 31 日在 Hugging Face 开源首个多模态模型 DeepSeek-V4-Flash-Vision-Exp，采用 MIT License，公开模型文件、Tokenizer、Prompt Encoding 参考实现及最小化 PyTorch 推理实现。","source":"IT之家（RSS）","sourceUrl":"https://www.ithome.com/0/996/637.htm","aiHotUrl":"https://aihot.virxact.com/items/cmth7tmq2067orodmh6g0sxie","publishedAt":"2026-08-31T11:35:24.000Z","category":"行业动态","score":72,"selected":true,"articleBody":["IT之家：https://www.ithome.com/ 8 月 31 日消息，刚刚，DeepSeek V4 首个多模态模型 DeepSeek-V4-Flash-Vision-Exp 在 Hugging Face 上线，采用 MIT License 开源。","公开内容包括模型文件、Tokenizer、Prompt Encoding 参考实现，以及最小化 PyTorch 推理实现，覆盖视觉编码器、Aligner、DFlash Attention、MoE、Hyper-Connections 与 DSpark 等核心模块。","据IT之家了解， DeepSeek-V4-Flash-Vision-Exp 是 DeepSeek V4 系列首款原生支持图片输入的视觉模型，官方明确标注为“Exp”实验版本，于 2026 年 8 月 21 日上线 DeepSeek API 平台。","该模型支持在文本之外输入图片，可以让模型描述图片、识别截图中的文字、分析图表等。支持的图片格式包括 JPEG、PNG、GIF、WebP。","深度求索官方表示，V4-Flash-Vision-Exp 在各类 Agent 框架内展现出了较好的多模态适配能力，能够有效支持大家借助多样化的 Agent 工具解锁更多实用的工作场景。在 Agent、推理、世界知识等纯文本任务上与 V4-Flash 正式版持平，原有优势完整保留。在需要视觉理解的 Agent 基准测试中较 V4-Flash 大幅提升，多模态 Agent 能力已接近 Opus-4.8。"],"articleImages":[{"sourceUrl":"https://img.ithome.com/newsuploadfiles/2026/8/7cbaf458-6c4b-47a3-9054-c4945972eb6c.png?x-bce-process=image/format,f_auto","alt":"","afterParagraph":0,"url":"/media/articles/cmth7tmq2067orodmh6g0sxie/95bd6fc7043a456c.png"}],"mediaStatus":"ok","articleBodyZh":["DeepSeek-V4-Flash-Vision-Exp 模型已开源，多模态 Agent 能力接近 Opus-4.8 这条更新来自 ithome.com，发布时间为 2026-08-31，Aioga 保留原文入口以便核验。","摘要：DeepSeek 于 8 月 31 日在 Hugging Face 开源首个多模态模型 DeepSeek-V4-Flash-Vision-Exp，采用 MIT License，公开模型文件、Tokenizer、Prompt Encoding 参考实现及最小化 PyTorch 推理实现。","背景：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。","Aioga 观察：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。","影响与后续：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。 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。"],"translationStatus":"","bodyOrigin":"source-page","editorial":{"summary":"Hugging Face 已公开 DeepSeek V4 系列的首个视觉模型实验版本，发布内容覆盖权重、分词器、提示编码参考代码和基础推理实现，并采用 MIT License。","background":"该模型可接收 JPEG、PNG、GIF、WebP 图片，用于图片描述、截图文字识别和图表分析。来源称其于 2026 年 8 月 21 日上线 DeepSeek API 平台，并在视觉理解 Agent 基准中较 V4-Flash 有提升。","viewpoint":"Aioga 判断：此次发布不仅涉及模型文件，也包含视觉编码器、Aligner、MoE 等模块的最小化实现，便于外部开发者围绕其多模态处理方式进行测试；但性能结论主要来自官方表述。","implications":"可能影响：MIT License 与配套实现可能增加该模型被试用和集成的空间；但接近 Opus-4.8 的说法不代表所有任务都具备同等表现，实际效果需要结合独立评测、部署条件和具体 Agent 框架验证。","nextStep":"后续观察：应关注开源文件和推理实现能否顺利运行、不同视觉任务的复现结果，以及实验版本与 V4-Flash 正式版在文本、推理和 Agent 任务上的实际差异。","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":"2026-08-31T13:29:06.145Z","sourceHash":"209a0fbd4228119d","review":{"approved":true,"groundedness":95,"clarity":92,"duplicationRisk":12,"blockingIssues":[],"notes":["“权重”可改为来源原文的“模型文件”，以避免将公开模型文件进一步限定为权重。","“提示编码参考代码”可改为“Prompt Encoding 参考实现”，与来源表述更一致。","“便于外部开发者……”和“可能增加……空间”属于合理推断，已通过“判断”“可能”等措辞与来源事实区分。"]},"validation":{"passed":true,"mode":"ai-auto","revisions":1,"checks":["schema","length","source-attribution","editorial-labels","inference-boundary","low-source-overlap","no-html","independent-ai-review"]}},"tags":["行业动态","IT之家（RSS）"],"translations":{"zh-CN":{"title":"DeepSeek-V4-Flash-Vision-Exp 模型已开源，多模态 Agent 能力接近 Opus-4.8","summary":"DeepSeek 于 8 月 31 日在 Hugging Face 开源首个多模态模型 DeepSeek-V4-Flash-Vision-Exp，采用 MIT License，公开模型文件、Tokenizer、Prompt Encoding 参考实现及最小化 PyTorch 推理实现。","category":"行业动态","source":"IT之家","aggregationSource":"IT之家（RSS）","pageTitle":"DeepSeek-V4-Flash-Vision-Exp 模型已开源，多模态 Agent 能力接近 Opus-4.8 - Aioga AI资讯","description":"DeepSeek 于 8 月 31 日在 Hugging Face 开源首个多模态模型 DeepSeek-V4-Flash-Vision-Exp，采用 MIT License，公开模型文件、Tokenizer、Prompt Encoding 参考实现及最小化 PyTorch 推理实现。","url":"https://www.aioga.com/news/cmth7tmq2067orodmh6g0sxie/","articleBody":["DeepSeek-V4-Flash-Vision-Exp 模型已开源，多模态 Agent 能力接近 Opus-4.8 这条更新来自 ithome.com，发布时间为 2026-08-31，Aioga 保留原文入口以便核验。","摘要：DeepSeek 于 8 月 31 日在 Hugging Face 开源首个多模态模型 DeepSeek-V4-Flash-Vision-Exp，采用 MIT License，公开模型文件、Tokenizer、Prompt Encoding 参考实现及最小化 PyTorch 推理实现。","背景：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。","Aioga 观察：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。","影响与后续：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。 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。"]},"en":{"title":"DeepSeek-V4-Flash-Vision-Exp Model Open-Sourced, Multimodal Agent Capability Close to Opus-4.8","summary":"On August 31, DeepSeek open-sourced its first multimodal model, DeepSeek-V4-Flash-Vision-Exp, on Hugging Face under the MIT License, releasing the model files, Tokenizer, Prompt Encoding reference implementation, and a minimal PyTorch inference implementation.","category":"Industry","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"DeepSeek-V4-Flash-Vision-Exp Model Open-Sourced, Multimodal Agent Capability Close to Opus-4.8 - Aioga AI News","description":"On August 31, DeepSeek open-sourced its first multimodal model, DeepSeek-V4-Flash-Vision-Exp, on Hugging Face under the MIT License, releasing the model files, Tokenizer, Prompt En...","url":"https://www.aioga.com/en/news/cmth7tmq2067orodmh6g0sxie/","contentTranslated":true,"sourceHash":"c9d3e3259f504d70","translatedAt":"2026-08-31T12:48:21.399Z"},"ja":{"title":"DeepSeek-V4-Flash-Vision-Expモデルはオープンソースであり、そのマルチモーダルエージェント機能はOpus-4.8に近いものです","summary":"8月31日、DeepSeekは最初のマルチモーダルモデルであるDeepSeek-V4-Flash-Vision-ExpをHugging Face上でMITライセンスを用いてオープンソース化し、モデルファイル、トークナイザー、プロンプトエンコーディング参照実装、最小化PyTorch推論実装を公開しました。","category":"業界動向","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"DeepSeek-V4-Flash-Vision-Expモデルはオープンソースであり、そのマルチモーダルエージェント機能はOpus-4.8に近いものです - Aioga AIニュース","description":"8月31日、DeepSeekは最初のマルチモーダルモデルであるDeepSeek-V4-Flash-Vision-ExpをHugging Face上でMITライセンスを用いてオープンソース化し、モデルファイル、トークナイザー、プロンプトエンコーディング参照実装、最小化PyTorch推論実装を公開しました。","url":"https://www.aioga.com/ja/news/cmth7tmq2067orodmh6g0sxie/","contentTranslated":true,"sourceHash":"c9d3e3259f504d70","translatedAt":"2026-08-31T12:48:23.963Z"},"ko":{"title":"DeepSeek-V4-Flash-Vision-Exp 모델은 오픈 소스이며, 멀티모달 에이전트 기능은 Opus-4.8에 접근합니다","summary":"8월 31일, DeepSeek은 첫 멀티모달 모델인 DeepSeek-V4-Flash-Vision-Exp를 Hugging Face에서 MIT 라이선스를 사용해 오픈 소스화했으며, 모델 파일, 토큰라이저, 프롬프트 인코딩 참조 구현, 최소화된 PyTorch 추론 구현을 공개했습니다.","category":"업계 동향","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"DeepSeek-V4-Flash-Vision-Exp 모델은 오픈 소스이며, 멀티모달 에이전트 기능은 Opus-4.8에 접근합니다 - Aioga AI 뉴스","description":"8월 31일, DeepSeek은 첫 멀티모달 모델인 DeepSeek-V4-Flash-Vision-Exp를 Hugging Face에서 MIT 라이선스를 사용해 오픈 소스화했으며, 모델 파일, 토큰라이저, 프롬프트 인코딩 참조 구현, 최소화된 PyTorch 추론 구현을 공개했습니다.","url":"https://www.aioga.com/ko/news/cmth7tmq2067orodmh6g0sxie/","contentTranslated":true,"sourceHash":"c9d3e3259f504d70","translatedAt":"2026-08-31T12:48:31.819Z"},"es":{"title":"El modelo DeepSeek-V4-Flash-Vision-Exp es de código abierto, y sus capacidades multimodales de agente se acercan a Opus-4.8","summary":"El 31 de agosto, DeepSeek abrió su primer modelo multimodal, DeepSeek-V4-Flash-Vision-Exp, en Hugging Face, utilizando la licencia MIT, y los archivos de modelo, tokenizadores, implementaciones de referencia de codificación por prompt y minimización de inferencias de PyTorch.","category":"Industria","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"El modelo DeepSeek-V4-Flash-Vision-Exp es de código abierto, y sus capacidades multimodales de agente se acercan a Opus-4.8 - Aioga Noticias de IA","description":"El 31 de agosto, DeepSeek abrió su primer modelo multimodal, DeepSeek-V4-Flash-Vision-Exp, en Hugging Face, utilizando la licencia MIT, y los archivos de modelo, tokenizadores, imp...","url":"https://www.aioga.com/es/news/cmth7tmq2067orodmh6g0sxie/","contentTranslated":true,"sourceHash":"c9d3e3259f504d70","translatedAt":"2026-08-31T12:48:33.003Z"},"fr":{"title":"Le modèle DeepSeek-V4-Flash-Vision-Exp est open source, et ses capacités d’agent multimodal approchent Opus-4.8","summary":"Le 31 août, DeepSeek a rendu son premier modèle multimodal en open source, DeepSeek-V4-Flash-Vision-Exp, sur Hugging Face, utilisant la licence MIT, et les fichiers de modèles, tokenizeurs, implémentations de références par encodage de prompts et implémentations d’inférence PyTorch minimisées.","category":"Industrie","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Le modèle DeepSeek-V4-Flash-Vision-Exp est open source, et ses capacités d’agent multimodal approchent Opus-4.8 - Aioga Actualités IA","description":"Le 31 août, DeepSeek a rendu son premier modèle multimodal en open source, DeepSeek-V4-Flash-Vision-Exp, sur Hugging Face, utilisant la licence MIT, et les fichiers de modèles, tok...","url":"https://www.aioga.com/fr/news/cmth7tmq2067orodmh6g0sxie/","contentTranslated":true,"sourceHash":"c9d3e3259f504d70","translatedAt":"2026-08-31T12:48:42.171Z"},"de":{"title":"Das DeepSeek-V4-Flash-Vision-Exp-Modell ist Open Source, und seine multimodalen Agentenfähigkeiten nähern sich Opus-4.8","summary":"Am 31. August veröffentlichte DeepSeek sein erstes multimodales Modell, DeepSeek-V4-Flash-Vision-Exp, auf Hugging Face als Open-Source, unter Verwendung der MIT-Lizenz, und veröffentlichte Modelldateien, Tokenizer, Prompt-Codierungs-Referenzimplementierungen und minimierte PyTorch-Inferenzimplementierungen.","category":"行业动态","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Das DeepSeek-V4-Flash-Vision-Exp-Modell ist Open Source, und seine multimodalen Agentenfähigkeiten nähern sich Opus-4.8 - Aioga KI-News","description":"Am 31. August veröffentlichte DeepSeek sein erstes multimodales Modell, DeepSeek-V4-Flash-Vision-Exp, auf Hugging Face als Open-Source, unter Verwendung der MIT-Lizenz, und veröffe...","url":"https://www.aioga.com/de/news/cmth7tmq2067orodmh6g0sxie/","contentTranslated":true,"sourceHash":"c9d3e3259f504d70","translatedAt":"2026-08-31T12:48:42.150Z"},"pt-BR":{"title":"O modelo DeepSeek-V4-Flash-Vision-Exp é open source, e suas capacidades multimodais de agente se aproximam do Opus-4.8","summary":"Em 31 de agosto, a DeepSeek disponibilizou seu primeiro modelo multimodal, DeepSeek-V4-Flash-Vision-Exp, no Hugging Face, usando a licença MIT, e arquivos de modelo, tokenizadores, implementações de referência de codificação por prompt e implementações minimizadas de inferência do PyTorch.","category":"行业动态","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"O modelo DeepSeek-V4-Flash-Vision-Exp é open source, e suas capacidades multimodais de agente se aproximam do Opus-4.8 - Aioga Notícias de IA","description":"Em 31 de agosto, a DeepSeek disponibilizou seu primeiro modelo multimodal, DeepSeek-V4-Flash-Vision-Exp, no Hugging Face, usando a licença MIT, e arquivos de modelo, tokenizadores,...","url":"https://www.aioga.com/pt-BR/news/cmth7tmq2067orodmh6g0sxie/","contentTranslated":true,"sourceHash":"c9d3e3259f504d70","translatedAt":"2026-08-31T12:48:50.530Z"},"ru":{"title":"Модель DeepSeek-V4-Flash-Vision-Exp является открытым исходным кодом, а её мультимодальные агентские возможности приближаются к Opus-4.8","summary":"31 августа DeepSeek открыла свою первую мультимодальную модель DeepSeek-V4-Flash-Vision-Exp на Hugging Face с использованием лицензии MIT, а также публично раскрытые файлы моделей, токенайзеры, референсные реализации кодирования prompt и минимизированные реализации вывода PyTorch.","category":"行业动态","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Модель DeepSeek-V4-Flash-Vision-Exp является открытым исходным кодом, а её мультимодальные агентские возможности приближаются к Opus-4.8 - Aioga Новости ИИ","description":"31 августа DeepSeek открыла свою первую мультимодальную модель DeepSeek-V4-Flash-Vision-Exp на Hugging Face с использованием лицензии MIT, а также публично раскрытые файлы моделей,...","url":"https://www.aioga.com/ru/news/cmth7tmq2067orodmh6g0sxie/","contentTranslated":true,"sourceHash":"c9d3e3259f504d70","translatedAt":"2026-08-31T12:48:51.272Z"},"ar":{"title":"نموذج DeepSeek-V4-Flash-Vision-Exp مفتوح المصدر، وقدراته متعددة الوسائط تقترب من Opus-4.8","summary":"في 31 أغسطس، قامت ديب سيك بفتح أول نموذج متعدد الوسائط لها، DeepSeek-V4-Flash-Vision-Exp، على تطبيق Hugging Face، باستخدام رخصة MIT، وكشف عن ملفات النماذج، والرمزيات، وتطبيقات الترميز المتكرر، وتقليل تنفيذ الاستنتاج في PyTorch.","category":"行业动态","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"نموذج DeepSeek-V4-Flash-Vision-Exp مفتوح المصدر، وقدراته متعددة الوسائط تقترب من Opus-4.8 - Aioga أخبار الذكاء الاصطناعي","description":"في 31 أغسطس، قامت ديب سيك بفتح أول نموذج متعدد الوسائط لها، DeepSeek-V4-Flash-Vision-Exp، على تطبيق Hugging Face، باستخدام رخصة MIT، وكشف عن ملفات النماذج، والرمزيات، وتطبيقات التر...","url":"https://www.aioga.com/ar/news/cmth7tmq2067orodmh6g0sxie/","contentTranslated":true,"sourceHash":"c9d3e3259f504d70","translatedAt":"2026-08-31T12:48:59.535Z"},"hi":{"title":"डीपसीक-वी4-फ्लैश-विजन-एक्सपी मॉडल ओपन सोर्स है, और इसकी मल्टीमॉडल एजेंट क्षमताएं ओपस-4.8 तक पहुंचती हैं","summary":"31 अगस्त को, डीपकसीक ने एमआईटी लाइसेंस का उपयोग करते हुए, हगिंग फेस पर अपना पहला मल्टीमॉडल मॉडल, डीपसीक-वी4-फ्लैश-विजन-एक्सपी ओपन-सोर्स किया, और सार्वजनिक रूप से प्रकट मॉडल फ़ाइलें, टोकनाइज़र, प्रॉम्प्ट एन्कोडिंग संदर्भ कार्यान्वयन और PyTorch अनुमान कार्यान्वयन को कम किया।","category":"行业动态","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"डीपसीक-वी4-फ्लैश-विजन-एक्सपी मॉडल ओपन सोर्स है, और इसकी मल्टीमॉडल एजेंट क्षमताएं ओपस-4.8 तक पहुंचती हैं - Aioga AI समाचार","description":"31 अगस्त को, डीपकसीक ने एमआईटी लाइसेंस का उपयोग करते हुए, हगिंग फेस पर अपना पहला मल्टीमॉडल मॉडल, डीपसीक-वी4-फ्लैश-विजन-एक्सपी ओपन-सोर्स किया, और सार्वजनिक रूप से प्रकट मॉडल फ़ाइलें...","url":"https://www.aioga.com/hi/news/cmth7tmq2067orodmh6g0sxie/","contentTranslated":true,"sourceHash":"c9d3e3259f504d70","translatedAt":"2026-08-31T12:49:00.334Z"},"it":{"title":"Il modello DeepSeek-V4-Flash-Vision-Exp è open source e le sue capacità di agente multimodale si avvicinano a Opus-4.8","summary":"Il 31 agosto, DeepSeek ha reso open source il suo primo modello multimodale, DeepSeek-V4-Flash-Vision-Exp, su Hugging Face, utilizzando la licenza MIT, e file di modello divulgati pubblicamente, tokenizatori, implementazioni di riferimento per codifica dei prompt e implementazioni di inferenza PyTorch minimizzate.","category":"行业动态","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Il modello DeepSeek-V4-Flash-Vision-Exp è open source e le sue capacità di agente multimodale si avvicinano a Opus-4.8 - Aioga Notizie IA","description":"Il 31 agosto, DeepSeek ha reso open source il suo primo modello multimodale, DeepSeek-V4-Flash-Vision-Exp, su Hugging Face, utilizzando la licenza MIT, e file di modello divulgati...","url":"https://www.aioga.com/it/news/cmth7tmq2067orodmh6g0sxie/","contentTranslated":true,"sourceHash":"c9d3e3259f504d70","translatedAt":"2026-08-31T12:49:09.380Z"},"nl":{"title":"Het DeepSeek-V4-Flash-Vision-Exp-model is open source en de multimodale agentmogelijkheden benaderen Opus-4.8","summary":"Op 31 augustus maakte DeepSeek zijn eerste multimodale model, DeepSeek-V4-Flash-Vision-Exp, open source op Hugging Face, met behulp van de MIT-licentie, en maakte modelbestanden, tokenizers, prompt-encoding referentie-implementaties en geminimaliseerde PyTorch-inferentie-implementaties openbaar.","category":"行业动态","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Het DeepSeek-V4-Flash-Vision-Exp-model is open source en de multimodale agentmogelijkheden benaderen Opus-4.8 - Aioga AI-nieuws","description":"Op 31 augustus maakte DeepSeek zijn eerste multimodale model, DeepSeek-V4-Flash-Vision-Exp, open source op Hugging Face, met behulp van de MIT-licentie, en maakte modelbestanden, t...","url":"https://www.aioga.com/nl/news/cmth7tmq2067orodmh6g0sxie/","contentTranslated":true,"sourceHash":"c9d3e3259f504d70","translatedAt":"2026-08-31T12:49:09.355Z"},"tr":{"title":"DeepSeek-V4-Flash-Vision-Exp modeli açık kaynaklıdır ve çoklu modal ajan yetenekleri Opus-4.8'e yaklaşır","summary":"31 Ağustos'ta DeepSeek, ilk çoklu modal modeli DeepSeek-V4-Flash-Vision-Exp'i MIT Lisansı kullanarak Hugging Face'te açık kaynak olarak tanıttı ve kamuya açık model dosyalarını, tokenizer'ları, prompt kodlama referans uygulamalarını ve en aza indirilmiş PyTorch çıkarım uygulamalarını açıkladı.","category":"行业动态","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"DeepSeek-V4-Flash-Vision-Exp modeli açık kaynaklıdır ve çoklu modal ajan yetenekleri Opus-4.8'e yaklaşır - Aioga AI Haberleri","description":"31 Ağustos'ta DeepSeek, ilk çoklu modal modeli DeepSeek-V4-Flash-Vision-Exp'i MIT Lisansı kullanarak Hugging Face'te açık kaynak olarak tanıttı ve kamuya açık model dosyalarını, to...","url":"https://www.aioga.com/tr/news/cmth7tmq2067orodmh6g0sxie/","contentTranslated":true,"sourceHash":"c9d3e3259f504d70","translatedAt":"2026-08-31T12:49:18.430Z"},"vi":{"title":"Mô hình DeepSeek-V4-Flash-Vision-Exp là mã nguồn mở, và khả năng tác nhân đa phương thức của nó gần với Opus-4.8","summary":"Vào ngày 31 tháng 8, DeepSeek đã mở mã nguồn mô hình đa phương thức đầu tiên của mình, DeepSeek-V4-Flash-Vision-Exp, trên Hugging Face, sử dụng Giấy phép MIT, đồng thời công khai các tệp mô hình, bộ mã hóa, các triển khai mã hóa nhắc nhở, và các triển khai suy luận PyTorch được tối giản.","category":"行业动态","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Mô hình DeepSeek-V4-Flash-Vision-Exp là mã nguồn mở, và khả năng tác nhân đa phương thức của nó gần với Opus-4.8 - Tin tức AI Aioga","description":"Vào ngày 31 tháng 8, DeepSeek đã mở mã nguồn mô hình đa phương thức đầu tiên của mình, DeepSeek-V4-Flash-Vision-Exp, trên Hugging Face, sử dụng Giấy phép MIT, đồng thời công khai c...","url":"https://www.aioga.com/vi/news/cmth7tmq2067orodmh6g0sxie/","contentTranslated":true,"sourceHash":"c9d3e3259f504d70","translatedAt":"2026-08-31T12:49:18.335Z"},"id":{"title":"Model DeepSeek-V4-Flash-Vision-Exp bersifat open source, dan kemampuan agen multimodalnya mendekati Opus-4.8","summary":"Pada 31 Agustus, DeepSeek membuka sumber model multimodal pertamanya, DeepSeek-V4-Flash-Vision-Exp, di Hugging Face, menggunakan Lisensi MIT, serta mengungkapkan file model secara publik, tokenizer, implementasi referensi prompt encoding, dan implementasi inferensi PyTorch yang diminimalkan.","category":"行业动态","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Model DeepSeek-V4-Flash-Vision-Exp bersifat open source, dan kemampuan agen multimodalnya mendekati Opus-4.8 - Berita AI Aioga","description":"Pada 31 Agustus, DeepSeek membuka sumber model multimodal pertamanya, DeepSeek-V4-Flash-Vision-Exp, di Hugging Face, menggunakan Lisensi MIT, serta mengungkapkan file model secara...","url":"https://www.aioga.com/id/news/cmth7tmq2067orodmh6g0sxie/","contentTranslated":true,"sourceHash":"c9d3e3259f504d70","translatedAt":"2026-08-31T12:49:27.361Z"},"th":{"title":"โมเดล DeepSeek-V4-Flash-Vision-Exp เป็นโอเพ่นซอร์ส และความสามารถของเอเจนต์หลายรูปแบบใกล้เคียงกับ Opus-4.8","summary":"เมื่อวันที่ 31 สิงหาคม DeepSeek ได้เปิดซอร์สโมเดลมัลติโมดัลตัวแรก DeepSeek-V4-Flash-Vision-Exp บน Hugging Face โดยใช้ใบอนุญาต MIT และเปิดเผยต่อสาธารณะ รวมถึงไฟล์โมเดล, tokenizer, การเข้ารหัสอ้างอิงแบบ prompt encoding และการลดจํานวนการใช้งาน PyTorch inference","category":"行业动态","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"โมเดล DeepSeek-V4-Flash-Vision-Exp เป็นโอเพ่นซอร์ส และความสามารถของเอเจนต์หลายรูปแบบใกล้เคียงกับ Opus-4.8 - ข่าว AI Aioga","description":"เมื่อวันที่ 31 สิงหาคม DeepSeek ได้เปิดซอร์สโมเดลมัลติโมดัลตัวแรก DeepSeek-V4-Flash-Vision-Exp บน Hugging Face โดยใช้ใบอนุญาต MIT และเปิดเผยต่อสาธารณะ รวมถึงไฟล์โมเดล, tokenizer, ก...","url":"https://www.aioga.com/th/news/cmth7tmq2067orodmh6g0sxie/","contentTranslated":true,"sourceHash":"c9d3e3259f504d70","translatedAt":"2026-08-31T12:49:27.494Z"},"pl":{"title":"Model DeepSeek-V4-Flash-Vision-Exp jest open source, a jego możliwości multimodalne agentów zbliżają się do Opus-4.8","summary":"31 sierpnia DeepSeek udostępnił swój pierwszy multimodalny model, DeepSeek-V4-Flash-Vision-Exp, na Hugging Face, korzystając z licencji MIT, i publicznie ujawnił pliki modelu, tokenizery, referencyjne implementacje prompt encoding oraz zminimalizowane implementacje wnioskowania PyTorch.","category":"行业动态","source":"IT之家（RSS）","aggregationSource":"IT之家（RSS）","pageTitle":"Model DeepSeek-V4-Flash-Vision-Exp jest open source, a jego możliwości multimodalne agentów zbliżają się do Opus-4.8 - Aioga Wiadomości AI","description":"31 sierpnia DeepSeek udostępnił swój pierwszy multimodalny model, DeepSeek-V4-Flash-Vision-Exp, na Hugging Face, korzystając z licencji MIT, i publicznie ujawnił pliki modelu, toke...","url":"https://www.aioga.com/pl/news/cmth7tmq2067orodmh6g0sxie/","contentTranslated":true,"sourceHash":"c9d3e3259f504d70","translatedAt":"2026-08-31T12:49:36.277Z"}},"evidenceTier":"verified-news","reviewStatus":"editorial-selected","indexable":true,"editorialCover":""}}