{"full_name":"Edge0-AI/Edge0","owner":"Edge0-AI","name":"Edge0","shard":"1","url":"https://github.com/Edge0-AI/Edge0","description":"Device-native foundation models: 145B on a laptop, 35B on a phone. Latent MoE, SSD-streamed weights, a router that predicts loads a step ahead. Mac, iPhone, Android, Windows.","homepage":"https://edge0.ai/","topics":["edge-ai","foundation-model","on-device-ai"],"tag_pages":["on-device-ai"],"language":"C++","license":"Apache-2.0","avatar":"https://avatars.githubusercontent.com/u/310437271?v=4","created_at":"2026-09-08T13:41:47Z","pushed_at":"2026-11-16T07:55:27Z","metrics":{"stars":11746,"forks":1856,"watchers":850,"issues":47,"prs":110,"releases":3,"issues_open":19,"issues_closed":28,"prs_open":34,"prs_closed":18,"prs_merged":58,"discussions":7},"growth":{"day":{"stars":95,"forks":26,"base_day":"2026-11-15"},"week":{"stars":2570,"forks":442,"base_day":"2026-11-09"}},"history":{"dates":["2026-11-09","2026-11-10","2026-11-11","2026-11-12","2026-11-13","2026-11-14","2026-11-15","2026-11-16"],"stars":[9176,9180,9697,10149,10686,11252,11651,11746],"forks":[1414,1414,1460,1548,1633,1732,1830,1856],"watchers":[634,634,665,710,710,756,802,850],"issues_open":[null,null,5,5,6,14,16,19],"prs_open":[null,null,28,29,29,24,28,34]},"boards":{"weekly":54,"monthly":199,"weekly_rate":46,"monthly_rate":116,"daily_forks":196,"daily_watchers":5,"new":36},"same_owner":[],"issues":[{"number":201,"title":"Support for Apple Silicon (Metal) for Edge0 120B / 145B","comments":5,"created_at":"2026-11-13T10:08:40Z"},{"number":197,"title":"Support for AMD GPUs (Vulkan or ROCm/HIP) for Edge0 120B / 145B","comments":5,"created_at":"2026-11-12T20:56:09Z"},{"number":205,"title":"[Engine] Layer split over two CUDA sm_86 cards: stage 2 never gets prompt rows, stall at layer 1 of a small tail chunk","comments":4,"created_at":"2026-11-13T13:12:43Z"},{"number":209,"title":"Docs and reachability feedback from a from-source Windows build (fb0c9b6)","comments":2,"created_at":"2026-11-13T18:10:19Z"},{"number":208,"title":"Performance report: edge0-145b / RTX 4090 24 GB / 64 GB RAM / Windows (from-source fb0c9b6)","comments":2,"created_at":"2026-11-13T18:10:18Z"},{"number":107,"title":"CUDA support timeline","comments":2,"created_at":"2026-09-16T23:05:47Z"},{"number":233,"title":"[Engine] Helper caches and --peer-device give wrong output on latent packs (Edge0-145B NVFP4, sm_86)","comments":1,"created_at":"2026-11-15T17:38:29Z"},{"number":217,"title":"Serving fails with ModuleNotFoundError: No module named 'jinja2.sandbox'","comments":1,"created_at":"2026-11-14T14:57:38Z"}],"discussions":[{"number":220,"title":"Edge0-145B Q3_K on an RTX 3090 (24 GB, PCIe Gen 2): community bench at 24 / 16 GiB, plus why 12 GiB doesn't fit","comments":0,"category":"Show and tell","created_at":"2026-11-14T21:30:53Z"},{"number":219,"title":"Edge0 145B Q3_K on Apple M5 Max (MacBook Pro, 128 GB): speed results on Metal","comments":0,"category":"Show and tell","created_at":"2026-11-14T20:16:41Z"},{"number":198,"title":"Withdrawn","comments":2,"category":"Show and tell","created_at":"2026-11-13T02:49:52Z"},{"number":195,"title":"distributed inference in a local cluster","comments":2,"category":"Ideas","created_at":"2026-11-12T17:32:20Z"},{"number":154,"title":"how to add model with the released Android app?","comments":1,"category":"General","created_at":"2026-10-28T06:45:26Z"},{"number":142,"title":"📊 Hardware Benchmarks — run edge0 on your device & share results","comments":1,"category":"General","created_at":"2026-10-15T09:23:40Z"},{"number":138,"title":"Welcome to edge0 Discussions!","comments":0,"category":"Announcements","created_at":"2026-10-15T07:59:53Z"}],"latest_release":{"name":"0.2.0","tag":"v0.2.0","published_at":"2026-11-11T15:13:28Z","body":"Windows and macOS. The Windows packages are built from `3203c39` on `main`, the macOS package from `f833336` on `main`. Installation and troubleshooting: see `README.md` inside each package.\n\n## What's Changed\n\n- **Edge0 145B and Edge0 120B.** Two latent-MoE models for a single NVIDIA GPU, run by edge0-strata.\n- **New inference engine, edge0-infer.** A single on-device MoE engine in C++ on GGML, with one native API, replaces the Python framework and the four separate platform apps. It includes a CLI (`run`, `chat`, `bench`, `serve`, `info`), an OpenAI-compatible local server and the Edge0 Chat desktop app. This release supports Edge0 35B; Edge0 8B follows in a later release. \n- **macOS (Apple Silicon).** Edge0 145B and Edge0 120B also run on Metal, and Edge0 Chat comes as a macOS package with the engines inside. (#214)\n- **Anthropic Messages API.** The local server also answers `POST /v1/messages` and `POST /v1/messages/count_tokens` for native models, and LAN authentication accepts `x-api-key`. (#180)\n\n## Windows x64\n\n`edge0-chat-0.2.0-windows-x64.zip`\n\n- Edge0 35B runs on the CPU (AVX2) or on a GPU through Vulkan.\n- Edge0 145B and Edge0 120B run on an NVIDIA GPU (CUDA) of compute capability 7.5 or newer (RTX 20 / 30 / 40 / 50) with at least 16 GiB of GPU memory, 20 GiB of system memory, and the NVIDIA CUDA Toolkit 13 installed.\n\nSHA256: `c7146b76d51c7986723b81a3643893e1ccbb09ce4f365a5090dc99e476bb31b9`\n\n## Windows Arm64\n\n`edge0-chat-0.2.0-windows-arm64.zip`\n\n- Edge0 35B runs on the CPU or on a GPU through Vulkan.\n- Edge0 145B and Edge0 120B run on an NVIDIA GPU (CUDA) of compute capability 12.1, such as the NVIDIA RTX Spark N1X, with at least 16 GiB of GPU memory, 20 GiB of system memory, and the NVIDIA CUDA Toolkit 13 for Windows Arm64 installed. The CUDA code is built for `sm_121` only.\n\nSHA256: `f64accb2c963058bc4ff1c9b064ccb35dc62ccdf8881b9a797d1ddab136f500a`\n\n## macOS (Apple Silicon)\n\n`edge0-chat-0.2.0-macos-arm64.zip`\n\n- macOS 26 or later on Apple Silicon. `edge0-infer` and edge0-strata are inside `Edge0 Chat.app`. The app is signed ad hoc and not notarized: on first start, allow it under **System Settings > Privacy & Security** (see `README.md` in the package).\n- Edge0 35B runs on Metal or the CPU.\n- Edge0 145B and Edge0 120B run on Metal from their `q3k` folders only; the NVFP4 folders do not run on macOS.\n\nSHA256: `4ece2c38302afd7b9610cd4cfadad9a82586c882d83b5eb3b31782efac6ea4e5`\n\n## Models\n\nModel weights are not included in the packages.\n\n- Edge0 145B: [Edge0-145B-NVFP4](https://huggingface.co/Edge0/Edge0-145B-NVFP4), [Edge0-145B-q3k](https://huggingface.co/Edge0/Edge0-145B-q3k)\n- Edge0 120B: [Edge0-120B-NVFP4](https://huggingface.co/Edge0/Edge0-120B-NVFP4), [Edge0-120B-q3k](https://huggingface.co/Edge0/Edge0-120B-q3k)\n\n**Full Changelog**: https://github.com/Edge0-AI/Edge0/compare/v0.1.1...v0.2.0\n","is_prerelease":false},"readme":true,"readme_zh":true,"news":[{"title":"A 35B language model running on an iPhone using only 1–2.5 GB of peak memory","url":"https://github.com/Edge0-AI/edge0/","hn":"https://news.ycombinator.com/item?id=49645864","source":"Hacker News","lang":"en","date":"2026-09-10","points":2,"comments":0},{"title":"Edge0 Releases Audio8-TTS-Realistic, a 0.6B Voice Cloning TTS Model","url":"https://www.marktechpost.com/2026/10/29/edge0-audio8-tts-realistic-0-6b-voice-cloning/","source":"MarkTechPost","lang":"en","date":"2026-10-28"},{"title":"搜 Edge0 却搜到 Aion：两个「Edge 系」端侧模型的辨认指南","url":"https://blog.csdn.net/gitblog_00269/article/details/168118223","source":"blog.csdn","lang":"zh","date":"2026-10-16"},{"title":"edge0 快速开始：5 分钟安装、下载并在 Mac 上跑通第一个 35B 模型对话","url":"https://blog.csdn.net/gitblog_00748/article/details/167119699","source":"blog.csdn","lang":"zh","date":"2026-10-05"},{"title":"Edge0-AI/Edge0: 今週のGitHubトレンドを詳しく解説!","url":"https://note.com/arti_sanal/n/n2694196dcf8a","source":"note","lang":"zh","date":"2026-10-05"},{"title":"Edge0-AI/Edge0: A Detailed Explanation of This Week's GitHub Trends!","url":"https://note.com/arti_sanal/n/n2694196dcf8a?hl=en","source":"note","lang":"en","date":"2026-10-04"},{"title":"Edge0 Framework Runs Qwen 35B AI Model from Your SSD","url":"https://www.geeky-gadgets.com/edge0-framework-ssd-ai/","source":"Geeky Gadgets","lang":"en","date":"2026-09-17"},{"title":"Edge0 Runs 35B AI Model on iPhone with Just 2.5 GB Memory","url":"https://x.com/i/trending/2097929280454107147","source":"x","lang":"en","date":"2026-09-09"},{"title":"Edge0-8b-a1b Preview","url":"https://huggingface.co/Edge0/Edge0-8B-A1B-preview","source":"huggingface","lang":"en","date":"2026-09-09"},{"title":"Run 35B Model on Phone Under 3GB Memory with Edge0","url":"https://news.google.com/rss/articles/CBMiXEFVX3lxTE9fU21aVjdVdllxUU94VW5zektyMUZDUnF5YlEzVzNKR1d3YjByOVR4Q3FFVGdCS1o4WG43VTRhOTg1dkx5eW1QSjZKT3RwbFlxZEVtMEFPa2x4dnpl?oc=5","source":"YouTube","lang":"en","date":"2026-09-11"}],"stale":false,"snapshot_at":"2026-11-16T09:23:14Z","history_start":"2026-09-06"}