AI News

Pi （Agent Harness） 是极简开源的终端编程 Agent，由 Mario Zechner 推出，同时是 OpenClaw 的底层 Coding Agent。工具只给模型 read、write、edit、bash ...

Orbis 是 Visko 推出的首个 Live Model（实时世界模型），让 AI 视频像直播一样持续运转。用户可随时输入新指令实时改变画面走向，无需等待成片。模型支持 4K...

MAI-Transcribe-2 是微软推出的最强AI语音转文字模型，在准确率、速度和成本上全面领先。模型支持60种语言，FLEURS测试集平均词错误率仅5.2%。

E-Commerce Bench 是阿里通义千问联合淘天集团开源的长程电商经营评测基准。E-Commerce Bench让 LLM Agent 以 10 万元本金在模拟真实市场中经营网店 365 天，...

GPT-6 Astra 是 OpenAI 推出的旗舰大模型，官方称其为「全球最智能、最符合人类意图的模型」。模型突破传统聊天边界，进化为能独立操作软件、浏览网页、完成...

Muse Spark 1.3 是 Meta 推出的新一代大语言模型，由扎克伯格亲自站台，定位为 Meta 在编程和智能体任务上的重大突破。

Fusion Model 是 FUMO Lab 推出的智能分配模型。模型能根据不断演化的任务状态，动态将计算需求分配给最适合的异构模型，实现'一次融合，尽揽众长'。

Gemini 3.8 是 Google 推出的推理与编码模型，包含 Flash 和 Flash Cyber 两个版本。Gemini 3.8 Flash 在保持与前代相同速度和低价的基础上，大幅提升软件工...

Easel 是浙江大学与北京大学联合开源的 AI 社交媒体内容工作台。平台基于 OpenClaw Agent 框架，用一个智能体打通社媒运营全链路，从热点发现、选题策划、内...

Claude Fable 5.1 is Anthropic's flagship large-scale model, positioned as one of the most powerful publicly available AIs. The model is specifically designed for complex reasoning, long-cycle agent tasks, and knowledge work, and is suitable for scientific research, programming, etc.

Claude Mythos 5.1 is Anthropic's flagship model in the Claude 5.1 series, sharing the same underlying capabilities as Claude Fable 5.1. The model is designed for high-risk professional use and is only available to approved online users...

Atlas, developed by World Labs founded by Fei-Fei Li, is the world's first multimodal world model. The model natively understands text, images, videos, and 3D spatial information, anchoring the scene in three-dimensional coordinates through spatial context...

Muse Voice Transcribe is Meta's first real-time audio-aware model, integrating streaming automatic speech recognition, speaker segmentation, and endpoint detection. The model supports over 70 languages and native code switching, and can process...

Hy4 preview lightweight version is a compressed version of the original 1.5TB Hy4 preview open-source model from Tencent Hunyuan, reduced to approximately 214GB using their self-developed Sherry 1.25bit sparse quantization and MIX-STQ1_0 layer-by-layer mixed precision technology...

Ling-3.0-flash-Fin is Ant Financial's first financial enhancement model, based on the Ling-3.0-flash architecture (124B total parameters/5.1B activation parameters) and designed for real financial workflows.

MiniMax H3 Max is a real-time video generation model from MiniMax, based on the open-source H3 platform with post-training and inference optimizations. The model supports text-to-image and image-to-video generation with simultaneous audio generation; a 5-second 768p video can be generated in less than 3 seconds...

Headlong is an open-source AI agent micro-framework from the Laude Institute, with a core of only about 10,000 lines of Bash script. Its biggest breakthrough is persistent agency; the agent never 'sleeps,' requiring no wake-up calls or scheduled tasks, functioning like a human...

Rome is an open-source Agentic OS. The tool enables AI to develop applications for users, allowing them to describe their needs in natural language. The agent automatically generates apps with user interfaces, data, and backend logic, which can be installed and run long-term...

Zing-0.5 is an open-source real-time interactive video world model from Loopit, with 5-byte parameters, based on the Wan2.2-TI2V-5B platform, and licensed under the Apache 2.0 license. Its core capability is 'joint control,' encompassing keyboard space manipulation and natural language processing...

Hy4 preview is Tencent's flagship open-source model, developed by Hunyuan, with a total of 770 bytes of parameters, 49 bytes of activation parameters, and a context length of 1MB. The model was developed through in-depth collaboration with experts from Tencent's software engineering, gaming, finance, and security fields...

shuohao-skills is an open-source collection of AI-powered short drama production skills, designed specifically for agents such as Claude Code and Codex. The tool provides services ranging from novel adaptation outlines, character design, art direction, scriptwriting to storyboarding...

Prime Agent is an open-source, long-lifecycle AI agent runtime framework from Prime Intellect. It uses RLM (Recursive Language Model) to variableize context and functionalize sub-agents, leveraging Continual Harness to achieve...

Gemini Omni 1.1 Flash is an AI video generation model launched by Google, aimed at developers and professional creators, supporting multi-level output from 360p quick preview to 4K ultra-high definition.

Hy-MT2-1.8B is a large-scale edge translation model launched by Tencent Hunyuan. With only 1.8B parameters, it supports mutual translation between 33 languages, and its translation quality surpasses commercial APIs such as Microsoft and Doubao on benchmarks such as FLORES-200.