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 | JustVugg/colibriRun frontier MoE models on hardware you already own — pure C, zero deps, experts streamed from disk. Tiny engine, immense model. 🐦 | C | +960 | 29,623 |
| 02 | sponsors/ever-co暂无简介 | TypeScript | +58 | 4,960 |
| 03 | bilawalsidhu/gods-eye-viewA spy satellite simulator in your browser, except the data is real. Live open source spatial intelligence on a photorealistic 3D globe. | JavaScript | +2,898 | 31,697 |
| 04 | tech-leads-club/agent-skillsThe secure, validated skill registry for professional AI coding agents. Extend Antigravity, Claude Code, Cursor, Copilot and more with absolute confidence. | TypeScript | +215 | 5,592 |
| 05 | melgarafael/DeskcommCRMOpen-source AI sales OS — self-hosted CRM with native AI agents + WhatsApp (WAHA). Open alternative to Kommo, Octadesk & Intercom for any business that sells by chat. MCP-ready, multi-tenant, LGPD. | TypeScript | +444 | 2,136 |
| 06 | sponsors/calesthioWorld's first open-source, agentic video production system. 12 production pipelines, 100+ tools, 700+ agent skill and production-knowledge files. Turn your AI coding assistant into a full video production studio. | Python | +383 | 58,353 |
| 07 | sponsors/asgeirtjExtracted system prompts from Anthropic - Claude Fable 5.1, Opus 5, Claude Design, Claude Code. OpenAI - ChatGPT GPT-6-Astra, Codex. Google - Gemini 3.8 Flash, 3.1 Pro, Antigravity. xAI - Grok, Grok Bot, Cursor, Kimi and more! Updated regularly. | JavaScript | +727 | 65,958 |
| 08 | vxcontrol/pentagiFully autonomous AI Agents system capable of performing complex penetration testing tasks | Go | +613 | 23,903 |
| 09 | multimodal-art-projection/YuEYuE2: frontier music generation with symbolic planning, zero-shot covers, and agentic music editing. | Python | +500 | 7,690 |
| 10 | yuliskov/SmartTubeBrowse media content with your own rules on Android TV | Java | +238 | 33,412 |
02 / MODELS
| 排名 | 模型 | 任务 | 收藏 | 下载 |
|---|---|---|---|---|
| 01 | Z-Image-Turbo | 文本生成图片 | 2,503 | 806,215 |
| 02 | MiniMax-H3 | — | 1,975 | 140,008 |
| 03 | Qwen3.8-27B | 视觉多模态理解 | 1,425 | 300,960 |
| 04 | DeepSeek-R1 | — | 1,381 | 927,325 |
| 05 | Qwen3.8-Flash-Next | 视觉多模态理解 | 1,167 | 51,157 |
| 06 | Qwen3-Coder-480B-A35B-Instruct | 文本生成 | 1,093 | 138,497 |
| 07 | IndexTTS-2 | 语音合成 | 977 | 207,739 |
| 08 | chatglm3-6b | — | 830 | 1,590,768 |
| 09 | Qwen3.8-2.4T-A95B | 文本生成 | 763 | 1,876 |
| 10 | Qwen-7B-Chat | 文本生成 | 757 | 693,267 |
DEVELOPER VIEW
colibri 值得动手试：零依赖意味着编译门槛极低，专家流式读取把 MoE 的显存需求转成了磁盘带宽需求，对只有消费级显卡或旧工作站的团队是实打实的可用增量。gods-eye-view 的价值在于数据是真的——它把开源空间情报做成了浏览器内可交互的产品形态，适合做地理可视化、城市分析类应用的原型底座。
agent-skills 解决的是真实痛点：技能包跨工具复用时的安全与版本问题，如果你的团队同时用多个编码助手，它提供了统一的注册与校验层。DeskcommCRM 对以聊天为成交入口的团队有直接替代价值：自托管、原生 Agent、WhatsApp 接入、多租户与 LGPD 合规，是 Kommo、Intercom 之外的开源选项。
模型侧，Qwen3.8 系列一次性覆盖视觉理解、Coder 与超大 MoE，选型时可以按上下文长度与部署形态分层取舍；IndexTTS-2 与 Z-Image-Turbo 则是当前语音与图像生成的性价比首选。
WORTH A STAR
colibri——它代表了一条被主流忽视的技术路线：不追算力，追工程。纯 C、零依赖、磁盘流式专家，这种"小引擎跑大模型"的思路一旦成熟，会直接改变本地部署的成本模型。
gods-eye-view——日增 2898 星不是偶然。真实数据 + 光追级 3D 地球 + 浏览器交付，把"空间智能"从论文词变成了可打开的产品，是本周最具想象力的可视化范式。
agent-skills——Agent 生态迟早需要一层可信的技能分发机制。它现在还很早期，但方向正确，且天然跨工具，适合提前跟进。
SOURCES & METHOD
GitHub Trending（项目热度、语言、Star） · ModelScope（模型任务、收藏与下载）
抓取日期：2026-09-14。排名是当日快照，指标可能随平台更新而变化。