Tag
Ant 发布了开源音频模型家族 Ming-3.0 Audio,采用 MIT 许可并开放权重与代码,支持语音与文本混合输入并可输出文字或语音回复,提供 126B 总参数的 Flash 和 9.3B 总参数的 tiny 两个 MoE 检查点。
littleguys 是一款开源桌面应用,以类 Whatsapp 的界面让用户管理 OpenClaws,支持浏览文件、操作浏览器、管理定时任务及媒体上传下载,并兼容 Hermes、Claude Code 和 Codex 等多个后端。
Helicon 是一个开源桌面应用和 VS Code 面板,为 Meta 的 Muse Code 编码代理 CLI 提供项目管理、会话侧边栏、审批界面和内联 diff 功能,支持 Windows、macOS、Linux 和 VS Code,免费且采用 MIT 许可。
OpenPilot is a free, MIT-licensed desktop AI agent that connects to any OpenAI-compatible model (OpenAI, OpenRouter, Ollama, LM Studio) to read and edit files and run commands in a chosen folder, with human approval for sensitive actions. Available on Windows now, with macOS in early beta.
美团开源了13.6B参数的LongCat-Video视频生成模型,原生支持文本到视频、图像到视频和视频续写,可在720p/30fps下生成数分钟长视频且无色彩衰减或质量下降,权重采用MIT许可证并允许商业使用。
ANUS 是一款开源编码代理,仅使用免费 LLM 运行:它维护一个免费模型排名列表,自动将请求路由到当前可用的最佳模型,当某个模型达到速率限制时自动切换到下一个备选模型。项目采用 MIT 协议,在 GitHub 上获得超过 6000 星。
Globestudio is a free, MIT-licensed open-source studio for creating dotted maps and 3D globes, offering 21 shader looks, customizable dots and gradients, and exports to PNG, SVG, WebM, MP4, GIF, or live embeds — with no account required.
The author developed DefaultBrowserChanger, a macOS menu bar app for switching browser profiles to isolate AI agent test sessions from daily browsing. It is an open-source, native Swift tool licensed under MIT.
The author built 'hunch', an open-source browser agent tool that uses a lightweight model (jev) for quick ~150ms decisions per step, reducing latency compared to LLM calls and integrating under existing agents like Claude Code with structured exit reports.
OpenVoice is an open-source AI tool for voice cloning that uses a short reference audio clip to replicate timbre, enabling multilingual and emotional adjustments, and has gained significant traction on GitHub.
Ling-3.0-flash-VL is a newly released multimodal AI model from inclusionAI, featuring native image and video understanding, a 124B parameter architecture with 5.5B active parameters, a 1M token context, and an MIT license.
The author recounts shifting their default software license from permissive MIT to the strong copyleft EUPL-1.2, arguing that permissive open-source licensing has mainly benefited large corporations at the expense of developers and users.
Harness Arena is an open-source blind benchmarking tool designed to compare various AI agent harnesses under controlled, anonymized conditions, aiming to provide fair evaluations and inviting community contributions.
The open-source workflow drama-skills distills a comic drama studio's experience into agent-readable skills for efficient AI short drama production, including scripts, character settings, storyboards, etc., and is open-sourced under the MIT license.
The tool called watermarks-remover has gained 16,000 stars in just 11 days, specifically designed to remove watermarks and traceability traces from AI-generated content, supporting various formats under the MIT open-source license.
Magiviz is an open-source AI video creation platform that provides a full-process AI workflow from writing scripts to generating videos, supporting up to 60-second videos with 12 built-in mainstream models, and the code is open-sourced under the MIT license.
The post details the quantization of Ornith 1.5 9B and 35B-A3B AI models using Atomic Dynamic methods, providing benchmarks against stock quantizations and sharing Hugging Face collections.
DeepSeek Harness is an open-source orchestration framework for AI coding agents with swappable plugins and sub-agent support, released with MIT license but accompanied by controversial API pricing changes.
The article introduces an open-sourced agent observability stack with four MIT-licensed repositories that focus on extracting and inspecting model beliefs to enhance AI agent debugging and performance.
A demo shows an open model generating a finished dashboard front end from a single prompt, but the underlying data is fake, prompting debate about whether the presentation layer is outpacing real reasoning.