#deepmind
4 articles
所有標籤claude-fable-5claude-mythos-5anthropicclaudemodel-releaseai-modelsgrok-imagine-videogrok-imagine-video-1-5xaiimage-to-videoai-videoseedance-2wan-2-7reve-2-0revegpt-image-2nano-banana-2nano-banana-proai-imageimage-generationviduvidu-q3video-apienterpriseb2bnvidiacosmos3cosmos3-nanoworld-modelphysical-airoboticsgemini-omniseedance-2-0kling-3-0video-generationmodel-comparisonflux-2imagen-4image-apikling-omnistoryboardingrunwaymodel-marketplacemulti-modelaudio-videotechnical-breakdownagnes-videoagnes-aipricingleaderboardclaude-sonnet-4-8leakgemini-3-5-flashgooglegoogle-ioagent-toolsdeepmindgemini-3-5gemini-3-5-progemini-omni-flashgemini-4claude-mythosgpt-5-5seedancebytedancegpt-5-6openaichatgptalignmentgeminiomniveohidreamopen-sourcediffusion-transformermultimodaltutorialrumorllmapicomparisonguidebestalternativeopenrouterawsazuregoogle-cloudinferenceimage-to-3d3d3d-generationtripo3dh3.1pbrquad-meshannouncementwavespeedaimultiview-to-3dtext-to-3dtext-to-imagenucleusnucleus-imageimage-to-imagematerialstexturepatinaimage-to-mapgame-devunrealunityblendermaterial-extractvideo-to-videovideo-inpaintingobject-removalmaskingvoidsam3text-to-audiomusic-generationmusic-coverstyle-transferminimaxpixverseanimationreference-to-videocharacter-consistencytext-to-videotransitionbaiduerniemultilingualchinesefastaudio-to-videomusic-videolip-syncrunway-mlimage-editingideogramsoraportrait-effectphoto-stylingparkouraction-videotalking-phototravel-photoad-generationspokespersonvirtual-try-onfashionai-avatarfree-toolsavatar-generatortalking-headimage-generatorvideo-generatorklingpredictionsdeevid-aiwanalibabavideo-editingvideo-editwan-2.7video-extendveo3start-end-to-videokuaishouelementsvacevideo-joinerwavespeed-aigemma-4on-device-aiaudio-converteraudio-processingfile-conversionimage-converterimage-processingface-blurprivacyvideo-convertervideo-processingai-tools4kmidjourneyfluxnano-bananaseedreambest-ai-image-generatorpixverse-v6audiovideo-effectsglmzhipu-aigptdeepseekai-newsphotaimage-enhanceupscalerimage-qualityphotorealisticcamera-controlvfxcybersecurityai-musicsunolyriamagihumandavincisand-aidigital-humandavinci-magihumanprofessionalai-image-generatorqwen-imagepollo-ailovartfreepikai-video-generatorbest-ai-video-generatorhiggsfieldkling-image-o3ai-image-generationgirl-filterface-transformationportraitsmile-filterphoto-editingwatermark-removalsora-alternativesora-shutdownpikagrokltxveo-4photo-colorizercolorizephoto-restorationbody-swapface-swapportrait-transferprismaudiovideo-to-audiofoleyai-audiosound-generationhunyuanaudio-generationv2aiclrrecraftrecraft-v4text-to-vectorsvgdesigndall-epeople-removerinpaintingfotorphoto-editorcontent-creationdesktop-appmp3wavflacaacpngjpgwebpheicmp4movaviwebmjanitor-aimedia-iovideo-editorm2.7ai-modelagentcodingbenchmarkage-filterentertainmentagingdog-selfiepet-contentgender-swapghibli-filteranimestudio-ghiblimidjourney-v8stable-diffusionbest-toolsai-content-detectorcontent-moderationcontent-safetynsfw-detectiontext-moderationimage-moderationvideo-moderationmoderation-apideveloper-guidesketch-to-videoinfinitetalkcelebrity-look-alikeface-recognitionclothes-changerfat-filtermemefortune-tellermath-solvereducationstory-generatorcreative-writingreviewbaseten2026canvafal-aifireworks-aileonardo-aimodalgpu-cloudreplicatecloudflarerunpodtogether-aiai-researchheliosbitdancebitdance-14bautoregressiveqwen-image-2typographyskyreelsskyreels-v3talking-avatarportrait-animationsoulxflashheadsoulx-flashheadreal-timestreamingai-imagessora-2q3wavespeed-desktopandroidmobileplaygroundbatch-processingloraworkflowai-pipelineffmpegaudio-conversionimage-conversionvideo-conversionvideo-mergevideo-trimmingvideo-enhancementvideo-upscaleinworldttstext-to-speechvoice-aicoming-soongpt-imagekimimoonshot-aiai-assistantlocal-aipersonal-aipredictiongenie-3interactive-environmentsmovaclawdbotpersonal-assistantautomationchatbotjavascripttypescriptsdkpythonspeculationai-collaborationproductivityai-agentsno-codeapp-builderdevelopmentapplebackground-removerface-enhancerimage-enhancementimage-eraserinpainttoolsclaude-codecodexai-codingcursordeveloper-toolsimage-enhancerai-platformshedraavatarsheygencreativevideo-marketingideasadobefireflyqualityrankingsimage-translationlocalizationimage-upscalingenhancementvideo-extensionvideo-upscalingdeveloperclipdropstability-aidalledeepaiperformanceblack-forest-labsvertex-aiinfrastructurehailuo-aihugging-facetencenttext-renderingimagenkling-ailuma-aidream-machineserverlessnightcafeai-artpika-labslm-arenadigital-twinstipsvideo-productionsynthesiadalle-3promptingavatarmulti-modalaimlapibytepluscomfyuidreaminakie-aiopenartpoyo-aiskyworktopazupscalingqwentrainingfine-tuningdepthcontrolnetposeupscaleoutpaintcannylightrickssdxlbackground-removaleventmarketinge-commerceproduct-photographymochicogvideosocial-mediainstagram
Gemini 3.5 Flash 正式發布——Flash 級模型在 Agent 基準測試上超越 Pro 級
Gemini 3.5 Flash 於 I/O 2026 正式推出，預設開啟思考模式，每百萬 token 定價 $1.50/$9，在 MCP Atlas 及大多數 Agent 測試套件上超越 Claude Opus 4.7 與 GPT-5.5。本文分析 Flash 的領先之處、落後之處，以及如何部署。
4 分鐘閱讀
Gemini 3.5 Pro 下個月即將到來——Flash 版本已透露的訊息
Google 在 I/O 2026 上發布了 Gemini 3.5 Flash，並將 Pro 版本推遲至六月。Flash 已在程式碼和代理基準測試上超越 Gemini 3.1 Pro，但在複雜推理方面有所退步——這正是 Pro 需要彌補的差距。以下是已知資訊、未知資訊，以及如何提前規劃。
3 分鐘閱讀
Google I/O 2026 的 Gemini 4.0：哪些已確認、哪些來自匿名消息、開發者真正需要關注什麼
Google I/O 今日上午 10 點（太平洋時間）正式開幕。關於新版 Gemini 的賽前報導從「漸進式 3.5 更新」到「深度整合的完整 Gemini 4.0」眾說紛紜。以下整理哪些是官方確認的資訊、哪些來自匿名消息來源，以及模型卡片發布後開發者應立即評估的七個面向。
2 分鐘閱讀
Google DeepMind Genie 3：創造互動環境的世界模型
探索Google DeepMind的Genie 3，這是一個突破性的世界模型，能從文字提示實時生成互動、可導航的虛擬環境。
1 分鐘閱讀