OPEN SOURCE OBSERVER
从项目热度走向工程判断：追踪开源工具、模型与开发者生态。
01 / CODE
| 排名 | 项目 | 语言 | 今日新增 | 总 Star |
|---|---|---|---|---|
| 01 | lnkiai/m3e-canvasSketch Material 3 Expressive screens in the browser and turn them into vibe-coding prompts. | TypeScript | — | 6,760 |
| 02 | sapientinc/PRAXISTAutonomous research system for measurable, computer-executable research. | Python | — | 6,756 |
| 03 | MengTo/threeuiOpen-source ThreeUI Community catalog with live interactive components and complete Community source. | HTML | — | 5,768 |
| 04 | s1dashu/ip-as-logo-skillA compact Agent Skill for highly simplified, rounded, subtly neo-skeuomorphic IP mascot logos. | — | — | 5,235 |
| 05 | CopilotKit/OpenBotOpen-source AI coworkers that each get a computer of their own: a browser, files and tools, with every action decided before it happens and recorded after. Bring any AG-UI agent. | TypeScript | — | 4,889 |
| 06 | amagine-ai/Amagine3DAmagine3D: From hardware requirements to editable 3D designs | Python | — | 4,675 |
| 07 | rakanki911/DLSS5-SwapperInstall, tune and restore DLSS 5 in your games with one click. Native RenoDX, DLSS5-Feeder for games without DLSS, OptiScaler and multipass routes; DirectX 8/9/11/12, Vulkan, OpenGL, DirectDraw and emulators. In-game F8 overlay, automatic backups, and a community page showing what works on your games and your graphics card. | JavaScript | — | 4,536 |
| 08 | XiaoDuoYa/codex-with-chatgptChatGPT thinks. Codex works. Use ChatGPT as the planning brain while keeping the Codex harness. | TypeScript | — | 4,294 |
| 09 | crmne/spotifastSpotify, native and fast. One lightweight Rust app for your whole library, local playback, and Spotify Connect on Linux, macOS, and Windows. | Rust | — | 4,055 |
| 10 | wang2122/sprix-sage-routerSprix AI at 屿智同行 — state-aware SELF/COLLABORATE/HANDOFF routing for A2A agent networks. | Python | — | 3,689 |
02 / MODELS
| 排名 | 模型 | 任务 | 收藏 | 下载 |
|---|---|---|---|---|
| 01 | Z-Image-Turbo | 文本生成图片 | 2,508 | 807,776 |
| 02 | MiniMax-H3 | — | 1,985 | 141,497 |
| 03 | Qwen3.8-27B | 视觉多模态理解 | 1,442 | 320,758 |
| 04 | DeepSeek-R1 | — | 1,381 | 927,340 |
| 05 | Qwen3.8-Flash-Next | 视觉多模态理解 | 1,181 | 52,999 |
| 06 | Qwen3-Coder-480B-A35B-Instruct | 文本生成 | 1,093 | 138,498 |
| 07 | IndexTTS-2 | 语音合成 | 979 | 207,976 |
| 08 | chatglm3-6b | — | 830 | 1,590,781 |
| 09 | Qwen3.8-2.4T-A95B | 文本生成 | 765 | 1,891 |
| 10 | Qwen-7B-Chat | 文本生成 | 757 | 693,273 |
DEVELOPER VIEW
免费可用的多模态能力：Qwen 系列和 Z-Image-Turbo 均可直接拉取，27B 级别视觉理解模型对个人开发者友好。
新范式参考：OpenBot 的"决策前置+动作留痕"设计，值得所有做 Agent 落地的团队抄作业。
性能向工具：spotifast 用 Rust 重写 Spotify 客户端并覆盖全平台，DLSS5-Swapper 提供一键调优与社区兼容性页面，都是典型的"官方没做，社区补上"。
代码生成专精：Qwen3-Coder-480B-A35B 是目前少见的开源超大代码模型，配合 codex-with-chatgpt 的规划分离思路，可以搭出低成本编码流水线。
WORTH A STAR
1. CopilotKit/OpenBot——"每个 AI 同事一台电脑"不是一个噱头，而是把权限、工具、审计三个难题打包解决的工程方案。它支持接入任意 AG-UI Agent，不锁定技术栈，对正在做 Agent 产品的团队参考价值极高。
2. sapientinc/PRAXIST——"可度量、计算机可执行的研究"这个定位很锋利。当前自主研究系统最大的问题是产出无法验证，PRAXIST 把可执行性当成第一约束，如果做扎实，会是 AI for Science 方向少见的务实派。
3. lnkiai/m3e-canvas——Material 3 Expressive 是 Google 新推的设计语言，社区工具还很少。用它画草图直接生成 vibe-coding 提示词，等于把设计意图一步送到编码 Agent 面前，适合独立开发者和设计工程师快速验证想法。
SOURCES & METHOD
GitHub 搜索 API（近 30 天新建 · 按 Star 排序）（项目热度、语言、Star） · ModelScope（模型任务、收藏与下载）
抓取日期：2026-09-15。排名是当日快照，指标可能随平台更新而变化。