[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"package:cask:llama-app:ru":3,"releases:stats:cask:llama-app:ru":252,"related:cask:llama-app:ru":253,"markdown:3:3796:1h0vqst":304},{"artifacts":4,"autoUpdates":45,"categories":46,"conflictsWith":52,"dependsOn":55,"deprecated":69,"description":70,"descriptionEn":71,"developer":72,"disabled":69,"displayName":73,"downloadSha256":74,"downloadSize":75,"downloadUrl":76,"editorChoice":69,"formulaeUrl":77,"homepage":78,"iconUrl":79,"installCommand":80,"installs":81,"installs30d":82,"isFont":69,"isLibrary":69,"kegOnly":69,"kind":85,"latestRelease":86,"machineTranslated":69,"minMacos":68,"name":73,"names":115,"platforms":118,"primaryCategory":223,"rank30d":224,"releaseCount":225,"releaseStats":226,"repoUrl":78,"screenshots":232,"sourceLocale":109,"sourceUrl":240,"summary":241,"summaryTranslation":242,"supports":243,"tags":245,"tap":250,"token":117,"version":114,"versionChangedAt":251},{"apps":5,"binaries":7,"entries":8,"pkgs":44},[6],"Llama.app",[],[9,16,23],{"declaration":10,"phase":14,"sources":15,"type":14},{"uninstall":11},[12],{"quit":13},"app.llama.Llama","uninstall",[],{"declaration":17,"phase":20,"sources":21,"target":19,"type":22},{"app":18,"target":19},[6],"\u002FApplications\u002FLlama.app","install",[6],"app",{"declaration":24,"phase":41,"sources":42,"type":43},{"zap":25},[26],{"trash":27},[28,29,30,31,32,33,34,35,36,37,38,39,40],"~\u002F.llama-app","~\u002F.local\u002Fbin\u002Fllama","~\u002FLibrary\u002FApplication Support\u002FLlama","~\u002FLibrary\u002FApplication Support\u002FLlamaBarn","~\u002FLibrary\u002FCaches\u002Fapp.llama.Llama","~\u002FLibrary\u002FCaches\u002Fapp.llamabarn.LlamaBarn","~\u002FLibrary\u002FCaches\u002FSentryCrash\u002FLlama","~\u002FLibrary\u002FHTTPStorages\u002Fapp.llama.Llama","~\u002FLibrary\u002FHTTPStorages\u002Fapp.llamabarn.LlamaBarn","~\u002FLibrary\u002FHTTPStorages\u002Fapp.llamabarn.LlamaBarn.binarycookies","~\u002FLibrary\u002FPreferences\u002Fapp.llama.Llama.plist","~\u002FLibrary\u002FPreferences\u002Fapp.llamabarn.LlamaBarn.plist","~\u002FLibrary\u002FWebKit\u002Fapp.llamabarn.LlamaBarn","cleanup",[],"zap",[],true,[47],{"icon":48,"machineTranslated":45,"name":49,"slug":50,"sourceLocale":51},"lucide:sparkles","Инструменты ИИ","ai","zh-CN",{"casks":53,"formulae":54},[],[],{"arch":56,"casks":58,"formulae":59,"macos":60,"requirements":61},[57],"arm64",[],[],">= 15",{"arch":62,"macos":66},[63],{"bits":64,"type":65},64,"arm",{">=":67},[68],"15",false,"Llama is a macOS menu-bar application for running local language models with llama.cpp. It suits developers and local-AI users who want a small controller, model downloads, a built-in chat interface and an API that other applications can use.\n\n## Models and API\nThe app starts a local server on port 9931, with API endpoints under \u002Fv1. It uses an installed llama.cpp engine or installs a prebuilt engine for the Mac. It discovers compatible existing models, recommends models that fit the machine and installs GGUF models from Hugging Face. Models load when requested and unload when idle. Use the built-in Web UI, compatible chat\u002Feditor clients, coding agents or direct API requests. Version 0.44.0 also supports decision models: the \u002Fv1\u002Fsystemone endpoint returns probabilities for typed-question options, and compatible models show a Decision chip. It can download available DFlash draft heads; the publisher reports up to 50% faster generation than MTP, which is a workload-dependent claim. A request-building page supports OpenAI chat, OpenAI Responses and Anthropic-style requests, with images, tools and structured-output settings where the selected model supports them.\n\n## Installation and resources\nDownload the official Llama DMG and put the app in Applications, or use `brew install --cask llama-app`. The current cask requires an Apple-silicon arm64 Mac and macOS 15 or later. The catalog now lists 0.44.0, while the current package download link still points to the historical 0.43.0 DMG (1,226,629 bytes). The official 0.44.0 Llama.dmg release asset is 1,257,897 bytes; these are distinct artifacts. Model weights and downloaded inference engines require much more disk space than this controller. Model size, quantization, context length and concurrency determine memory and performance requirements; the app's recommendations are useful but not a guarantee every model will run well.\n\n## Network, accounts and cost\nInference runs on the Mac. Downloading models or a prebuilt engine requires network access and is separate from local inference. Ordinary localhost use does not require a cloud inference subscription. Hugging Face repositories may be public or gated; their own access terms and model licenses apply. The app is MIT-licensed, while model weights and external tools have separate licenses. Tailscale access requires the separately installed and signed-in Tailscale service, with its own account and terms.\n\n## Network exposure and practical cautions\nBy default, the server is reachable only on the Mac. Settings can bind it to Tailscale or to the current network. The README warns that This network binds all interfaces and the server has no password by default; use that mode only on a trusted network, and avoid combining it with agent mode on a network you do not own. Local inference is not a promise that deliberately enabled network clients or downloads generate no traffic. Check who can reach the server and what connected agents can do.\n\nCustom model overrides in ~\u002F.config\u002Fllama\u002Fmodels.user.ini can override the app's memory-derived settings. Keep copies before changing them, and review ignored-option notices. External clients can supply private prompts or use tool-capable agents; check their permissions and generated actions. The inspected sources do not establish a blanket Accessibility or Full Disk Access requirement. Grant access only to the model\u002Fconfiguration locations and workflows you intend to use.\n\nSources: [Official project, storage and network behavior](https:\u002F\u002Fgithub.com\u002Fggml-org\u002FLlama-macOS), [0.44.0 release](https:\u002F\u002Fgithub.com\u002Fggml-org\u002FLlama-macOS\u002Freleases\u002Ftag\u002F0.44.0), [llama.cpp server API](https:\u002F\u002Fgithub.com\u002Fggml-org\u002Fllama.cpp\u002Ftree\u002Fmaster\u002Ftools\u002Fserver), [MIT license](https:\u002F\u002Fgithub.com\u002Fggml-org\u002FLlama-macOS\u002Fblob\u002Fmain\u002FLICENSE).","Menu bar app for running local LLMs","GGML organization and Llama contributors","Llama","f772824972442506614596879bb69192c4c8c17f4ead5d6bbb3ccdeef73f9b00",1257897,"https:\u002F\u002Fgithub.com\u002Fggml-org\u002FLlama-macOS\u002Freleases\u002Fdownload\u002F0.44.0\u002FLlama.dmg","https:\u002F\u002Fformulae.brew.sh\u002Fcask\u002Fllama-app","https:\u002F\u002Fgithub.com\u002Fggml-org\u002FLlama-macOS","https:\u002F\u002Fcdn.opennavo.com\u002Ficons\u002Fuploads\u002F8fd3bf680130\u002Fa584f25a6670-256.png","brew install --cask llama-app",{"d30":82,"d365":83,"d90":84},280,1160,915,"cask",{"bodyMarkdown":87,"brewCommittedAt":88,"hasNotes":45,"id":89,"isLatest":45,"isPrerelease":69,"machineTranslated":69,"publishedAt":90,"sections":91,"source":108,"sourceLocale":109,"summary":110,"title":111,"translation":112,"version":114},"Llama now runs decision models. Instead of chatting, they answer a typed question with a probability for each option, through the server's `\u002Fv1\u002Fsystemone` endpoint. They carry a Decision chip, and their model page links to a guide on how to use them.\n\n- Download DFlash draft heads when available: up to 50% faster than MTP\n- Update the engine to b11429, switching right away when no model is loaded\n- Merge the Downloads and Command settings tabs into a new Advanced tab\n- Rename the Web UI settings tab to Chat and its setting to custom web app\n- Open the quick prompt without bringing Settings to the front\n\nSource: [Official 0.44.0 release](https:\u002F\u002Fgithub.com\u002Fggml-org\u002FLlama-macOS\u002Freleases\u002Ftag\u002F0.44.0).","2026-10-08T13:45:25Z",3427,"0001-01-01T00:00:00Z",[92,98,102],{"area":93,"items":94},"Models",[95,96,97],"Run decision models through \u002Fv1\u002Fsystemone with option probabilities.","Show a Decision chip and a guide link on compatible model pages.","Download available DFlash draft heads; publisher claims up to 50% over MTP.",{"area":99,"items":100},"Engine",[101],"Update to b11429; switch immediately when no model is loaded.",{"area":103,"items":104},"Settings",[105,106,107],"Merge Downloads and Command into Advanced.","Rename Web UI to Chat and the setting to custom web app.","Open the quick prompt without bringing Settings to the front.","editorial","en-US","Adds decision models and DFlash draft heads, and reorganizes settings.","Llama 0.44.0",{"locale":109,"machineTranslated":69,"sourceLocale":109,"status":113},"source","0.44.0",[73,116,117],"llamabarn","llama-app",[119,154,189],{"arch":57,"artifacts":120,"conflictsWith":140,"dependsOn":143,"downloadSha256":74,"downloadUrl":76,"macos":152,"minMacos":68,"requiresRosetta":69,"tag":153,"version":114},{"apps":121,"binaries":122,"entries":123,"pkgs":139},[6],[],[124,129,133],{"declaration":125,"phase":14,"sources":128,"type":14},{"uninstall":126},[127],{"quit":13},[],{"declaration":130,"phase":20,"sources":132,"target":19,"type":22},{"app":131,"target":19},[6],[6],{"declaration":134,"phase":41,"sources":138,"type":43},{"zap":135},[136],{"trash":137},[28,29,30,31,32,33,34,35,36,37,38,39,40],[],[],{"casks":141,"formulae":142},[],[],{"arch":144,"casks":145,"formulae":146,"macos":60,"requirements":147},[57],[],[],{"arch":148,"macos":150},[149],{"bits":64,"type":65},{">=":151},[68],"27","arm64_golden_gate",{"arch":57,"artifacts":155,"conflictsWith":175,"dependsOn":178,"downloadSha256":74,"downloadUrl":76,"macos":187,"minMacos":68,"requiresRosetta":69,"tag":188,"version":114},{"apps":156,"binaries":157,"entries":158,"pkgs":174},[6],[],[159,164,168],{"declaration":160,"phase":14,"sources":163,"type":14},{"uninstall":161},[162],{"quit":13},[],{"declaration":165,"phase":20,"sources":167,"target":19,"type":22},{"app":166,"target":19},[6],[6],{"declaration":169,"phase":41,"sources":173,"type":43},{"zap":170},[171],{"trash":172},[28,29,30,31,32,33,34,35,36,37,38,39,40],[],[],{"casks":176,"formulae":177},[],[],{"arch":179,"casks":180,"formulae":181,"macos":60,"requirements":182},[57],[],[],{"arch":183,"macos":185},[184],{"bits":64,"type":65},{">=":186},[68],"26","arm64_tahoe",{"arch":57,"artifacts":190,"conflictsWith":210,"dependsOn":213,"downloadSha256":74,"downloadUrl":76,"macos":68,"minMacos":68,"requiresRosetta":69,"tag":222,"version":114},{"apps":191,"binaries":192,"entries":193,"pkgs":209},[6],[],[194,199,203],{"declaration":195,"phase":14,"sources":198,"type":14},{"uninstall":196},[197],{"quit":13},[],{"declaration":200,"phase":20,"sources":202,"target":19,"type":22},{"app":201,"target":19},[6],[6],{"declaration":204,"phase":41,"sources":208,"type":43},{"zap":205},[206],{"trash":207},[28,29,30,31,32,33,34,35,36,37,38,39,40],[],[],{"casks":211,"formulae":212},[],[],{"arch":214,"casks":215,"formulae":216,"macos":60,"requirements":217},[57],[],[],{"arch":218,"macos":220},[219],{"bits":64,"type":65},{">=":221},[68],"arm64_sequoia",{"icon":48,"machineTranslated":45,"name":49,"slug":50,"sourceLocale":51},404,13,{"brewLag":227,"cadence":231,"count30d":229},{"compared":228,"earlierCount":229,"medianMinutes":230},4,1,153722867,"weekly",[233],{"caption":234,"height":235,"machineTranslated":45,"sourceLocale":109,"theme":236,"thumbUrl":237,"url":238,"width":239},"Исторический снимок строки меню версии 0.28 из официального README с названием LlamaBarn, показывающий установленные модели и настройки.",720,"light","https:\u002F\u002Fcdn.opennavo.com\u002Fscreenshots\u002Fuploads\u002Fscreenshot\u002Fe36068d98d10\u002Fe7ceab1f0a06-640.png","https:\u002F\u002Fcdn.opennavo.com\u002Fscreenshots\u002Fuploads\u002Fscreenshot\u002Fe36068d98d10\u002Fe7ceab1f0a06-1280.png",1280,"https:\u002F\u002Fgithub.com\u002Fhomebrew\u002Fhomebrew-cask\u002Fblob\u002FHEAD\u002FCasks\u002Fl\u002Fllama-app.rb","A macOS menu-bar controller for local GGUF models, llama.cpp inference, built-in web chat and compatible model APIs.",{"locale":109,"machineTranslated":69,"sourceLocale":109,"status":113},{"arm64":45,"requiresRosetta":69,"status":244,"x86_64":69},"known",[246,247,248,249],"Local AI","LLM","llama.cpp","Open source","homebrew\u002Fcask","2026-10-09T07:15:26.193153Z",null,[254,264,275,285,293],{"accentColor":255,"autoUpdates":69,"deprecated":69,"disabled":69,"displayName":256,"editorChoice":69,"iconUrl":257,"installs30d":258,"isFont":69,"isLibrary":69,"kind":85,"machineTranslated":45,"name":256,"primaryCategory":259,"rank30d":260,"sourceLocale":109,"summary":261,"token":262,"version":263,"versionChangedAt":251},"#18449E","Osaurus","https:\u002F\u002Fcdn.opennavo.com\u002Ficons\u002Fuploads\u002F27e60dc1e65e\u002F4b631a582622-256.png",134,{"icon":48,"machineTranslated":45,"name":49,"slug":50,"sourceLocale":51},658,"Нативное ИИ-приложение для Apple Silicon с локальными и облачными моделями, постоянными агентами, памятью, инструментами в изолированной среде и совместимыми API.","osaurus","0.25.20",{"accentColor":265,"autoUpdates":45,"deprecated":69,"disabled":69,"displayName":266,"editorChoice":69,"iconUrl":267,"installs30d":268,"isFont":69,"isLibrary":69,"kind":85,"machineTranslated":45,"name":266,"primaryCategory":269,"rank30d":270,"sourceLocale":109,"summary":271,"token":272,"version":273,"versionChangedAt":274},"#E7CC0D","EXO","https:\u002F\u002Fcdn.opennavo.com\u002Ficons\u002Fuploads\u002F64c002ff065e\u002F6112dd44ccff-256.png",227,{"icon":48,"machineTranslated":45,"name":49,"slug":50,"sourceLocale":51},468,"Локальная система ИИ-инференса с открытым исходным кодом, которая обнаруживает устройства поблизости и распределяет поддерживаемые модели по кластеру Mac.","exo","1.0.71","2026-10-07T09:38:28.442235Z",{"autoUpdates":45,"deprecated":69,"disabled":69,"displayName":276,"editorChoice":69,"iconUrl":277,"installs30d":278,"isFont":69,"isLibrary":69,"kind":85,"machineTranslated":45,"name":276,"primaryCategory":279,"rank30d":280,"sourceLocale":109,"summary":281,"token":282,"version":283,"versionChangedAt":284},"Perplexity AI","https:\u002F\u002Fcdn.opennavo.com\u002Ficons\u002Fuploads\u002F86b5e9494fcd\u002F8b68a118fae2-256.png",246,{"icon":48,"machineTranslated":45,"name":49,"slug":50,"sourceLocale":51},434,"Поиск и исследования с ИИ для Mac с рабочими процессами персонального компьютера, охватывающими подключённые файлы, нативные приложения и веб.","perplexity","26.39.0","2026-10-07T09:38:40.131182Z",{"autoUpdates":45,"deprecated":69,"disabled":69,"displayName":286,"editorChoice":69,"installs30d":287,"isFont":69,"isLibrary":69,"kind":85,"machineTranslated":69,"name":286,"primaryCategory":288,"rank30d":289,"sourceLocale":109,"summary":290,"token":291,"version":292,"versionChangedAt":274},"DeepL",166,{"icon":48,"machineTranslated":45,"name":49,"slug":50,"sourceLocale":51},583,"Приложение DeepL для Mac: перевод и улучшение текстов, перевод документов, глоссарии, сочетания клавиш и захват текста с экрана.","deepl","26.6.14916780",{"accentColor":294,"autoUpdates":45,"deprecated":69,"disabled":69,"displayName":295,"editorChoice":69,"iconUrl":296,"installs30d":297,"isFont":69,"isLibrary":69,"kind":85,"machineTranslated":69,"name":295,"primaryCategory":298,"rank30d":299,"sourceLocale":109,"summary":300,"token":301,"version":302,"versionChangedAt":303},"#FFCB28","Jan","https:\u002F\u002Fcdn.opennavo.com\u002Ficons\u002Fuploads\u002F959cc9db27d5\u002Faba6ce13c525-256.png",206,{"icon":48,"machineTranslated":45,"name":49,"slug":50,"sourceLocale":51},499,"Локальное ИИ-приложение: загрузка моделей, чат, ассистенты, облачные провайдеры, инструменты MCP и локальный API, совместимый с OpenAI.","jan","0.8.6","2026-10-09T11:16:07.323276Z","\u003Cp>Llama is a macOS menu-bar application for running local language models with llama.cpp. It suits developers and local-AI users who want a small controller, model downloads, a built-in chat interface and an API that other applications can use.\u003C\u002Fp>\n\u003Ch3 class=\"on-md-heading-1\">Models and API\u003C\u002Fh3>\n\u003Cp>The app starts a local server on port 9931, with API endpoints under \u002Fv1. It uses an installed llama.cpp engine or installs a prebuilt engine for the Mac. It discovers compatible existing models, recommends models that fit the machine and installs GGUF models from Hugging Face. Models load when requested and unload when idle. Use the built-in Web UI, compatible chat\u002Feditor clients, coding agents or direct API requests. Version 0.44.0 also supports decision models: the \u002Fv1\u002Fsystemone endpoint returns probabilities for typed-question options, and compatible models show a Decision chip. It can download available DFlash draft heads; the publisher reports up to 50% faster generation than MTP, which is a workload-dependent claim. A request-building page supports OpenAI chat, OpenAI Responses and Anthropic-style requests, with images, tools and structured-output settings where the selected model supports them.\u003C\u002Fp>\n\u003Ch3 class=\"on-md-heading-1\">Installation and resources\u003C\u002Fh3>\n\u003Cp>Download the official Llama DMG and put the app in Applications, or use \u003Ccode>brew install --cask llama-app\u003C\u002Fcode>. The current cask requires an Apple-silicon arm64 Mac and macOS 15 or later. The catalog now lists 0.44.0, while the current package download link still points to the historical 0.43.0 DMG (1,226,629 bytes). The official 0.44.0 Llama.dmg release asset is 1,257,897 bytes; these are distinct artifacts. Model weights and downloaded inference engines require much more disk space than this controller. Model size, quantization, context length and concurrency determine memory and performance requirements; the app's recommendations are useful but not a guarantee every model will run well.\u003C\u002Fp>\n\u003Ch3 class=\"on-md-heading-1\">Network, accounts and cost\u003C\u002Fh3>\n\u003Cp>Inference runs on the Mac. Downloading models or a prebuilt engine requires network access and is separate from local inference. Ordinary localhost use does not require a cloud inference subscription. Hugging Face repositories may be public or gated; their own access terms and model licenses apply. The app is MIT-licensed, while model weights and external tools have separate licenses. Tailscale access requires the separately installed and signed-in Tailscale service, with its own account and terms.\u003C\u002Fp>\n\u003Ch3 class=\"on-md-heading-1\">Network exposure and practical cautions\u003C\u002Fh3>\n\u003Cp>By default, the server is reachable only on the Mac. Settings can bind it to Tailscale or to the current network. The README warns that This network binds all interfaces and the server has no password by default; use that mode only on a trusted network, and avoid combining it with agent mode on a network you do not own. Local inference is not a promise that deliberately enabled network clients or downloads generate no traffic. Check who can reach the server and what connected agents can do.\u003C\u002Fp>\n\u003Cp>Custom model overrides in ~\u002F.config\u002Fllama\u002Fmodels.user.ini can override the app's memory-derived settings. Keep copies before changing them, and review ignored-option notices. External clients can supply private prompts or use tool-capable agents; check their permissions and generated actions. The inspected sources do not establish a blanket Accessibility or Full Disk Access requirement. Grant access only to the model\u002Fconfiguration locations and workflows you intend to use.\u003C\u002Fp>\n\u003Cp>Sources: \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fggml-org\u002FLlama-macOS\" target=\"_blank\" rel=\"noopener noreferrer\">Official project, storage and network behavior\u003C\u002Fa>, \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fggml-org\u002FLlama-macOS\u002Freleases\u002Ftag\u002F0.44.0\" target=\"_blank\" rel=\"noopener noreferrer\">0.44.0 release\u003C\u002Fa>, \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fggml-org\u002Fllama.cpp\u002Ftree\u002Fmaster\u002Ftools\u002Fserver\" target=\"_blank\" rel=\"noopener noreferrer\">llama.cpp server API\u003C\u002Fa>, \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fggml-org\u002FLlama-macOS\u002Fblob\u002Fmain\u002FLICENSE\" target=\"_blank\" rel=\"noopener noreferrer\">MIT license\u003C\u002Fa>.\u003C\u002Fp>\n"]