**A next-generation family of agentic models built for long-horizon tasks in real-world environments.

Async Python wrapper for Seedance, Kling, MiniMax and Wan video generation. Supports CLI and a local web UI

[`Blog`] [`Code`] [`🤗 Weights`] [`ModelScope`] [`License`] We release **VDN-Minimax-H3** (**VDN-H3**), a hybrid-attention model that generates video faster than it plays, powered by MiniMax H3.

Track and visualize the stars history of any GitHub repository. Open-source growth analytics and velocity tracking.

A cPanel and WHM authentication bypassing tool

text-humanizer is an open-source project designed to convert text generated by AI to its "humanized" version. Bypasses the most of AI detectors such as Turnitin or GPTZero

Heretic v1.2.0+custom with the Arbitrary-Rank Ablation (ARA) method

**4-bit dynamic quant.

Audit your GitHub repository search ranking signals. Checks name, description, topics, README, stars, forks, and activity.

Computer use for about $0.0002 a step: OCR the screen, classify the next action with TypeSafe, click. macOS.

Can we run something like Jev on a 3090 at home?

A fruit fly connectome as a drone pilot. Camera -> fly eyes -> spiking brain -> descending neurons -> drone. Live 3D demo in your browser.

a list of awesome open source android apps 📱

  

A small neural network that decides whether a number is Numberwang.

小红书API、小红书接口：笔记搜索、笔记详情、评论查询的 Python 调用示例。Just One API。

image2.5 (GPT-Image-2.5) API gateway on APIMart: Flare and Sunburst model IDs, the $0.0085-per-image relay route, OpenAI-compatible endpoints and a 12-image prompt gallery.

Train and chat with a frozen language model coupled to the full retained MaleCNS fly connectome.

A TypeSafe/Jev agent that plays Super Mario Bros. from structured emulator state.

Add a decorator for dspy Signatures that automatically uses TypeSafe where relevant

蓝色星原：旅谣自动化脚本

Command-line Cloudflare Turnstile token generator: discover the sitekey on any page and print a cf-turnstile-response token for your scripts, tests and CI.

**8-bit dynamic quant.

![Python 3.

A Qualifying Exam for Coding Agent as Robot Learning Engineers

AIBuildAI LLM-Post-Train Agent: a recursive self-improving (RSI) agent for autonomous LLM post-training

EvoOntology: A Self-Evolving Ontology Layer for Data Agents ⚙️ EvoOntology插件为Claude Code/Codex 建立&进化本体层

Akahya is a Turkish text-to-speech adaptation of OpenBMB's VoxCPM2, released by Alican Kiraz.

KnoxMap: turn any real place into a playable Project Zomboid Build 42 map (built on Knoxify)

My Python learning journey — basic programs, practice, and projects.

SmuggleMyPayload generates HTML pages that embed and reconstruct files client-side using JavaScript. The toolkit offers multiple HTML Smuggling techniques, payload encoding options, JavaScript variations, and pre-built HTML templates.

Local-first detection and temporal localization of reused footage in egocentric video.

[SIGGRAPH Asia 2026] CustomDance: Customized 3D Dance Generation with Coarse-to-Fine Human-Centered Interactive Control

灯暗一点，漫画翻慢一点。支持页内悄悄话、点赞分享与按需看图的独立共读放映厅。

GPU-native quality gate for first-person video

Experimental ComfyUI SelfLift variant for MiniMax H3 video/audio masks and audio preservation.

Enumerate user accounts and registered authentication methods via the Microsoft Self-Service Password Reset (SSPR) portal

A GPT-6 Astra agent that runs a live board of every Robinhood Stock Token against its real price and calls the gap while the real NYSE is closed. Built with Minara.

Twelve image2.5 (GPT-Image-2.5) prompt recipes with example outputs, model version, aspect ratio and copy-paste cURL, Python and JavaScript calls through the APIMart API.

从「龙猫体育锻炼」微信小程序自己已登录的进程里读出 token 的本地命令行工具

FS_Audio Suite: modular YuE2 audio generation nodes for ComfyUI, by Make the Robot Do It

**Dynamic 4-bit quant with 8-bit attention.

Camera-only autonomous drone in MuJoCo with a small judgment model (TypeSafe Jev) in the loop at 2.5Hz

## Full face muscle dynamics with clear English speech *V2 is out! features multi-gender male/female voice with flawless speech sync and crystal clear quality*

SHADOW 50M Instruct: a proof of concept. 44M ternary parameters, exact circuits inside the model, memory on disk, 19.8 MB, offline on a CPU.

谁说AI不能生成GIF? 本skill利用GPT-Image 2.5 一次性生成多格帧图 - 合成为动画gif - 一次调用就生成动态图。产品动图生成器 > 你给我产品照片，我还你一段会自动循环播放的产品动图。 这个技能是给"不会用设计软件、但需要动图素材"的人用的。 你不需要懂任何技术，只要会发图片、会说话就够了。

让 Windows 微信账号活起来：支持任意 OpenAI 兼容模型的 AI 群聊机器人

## The Complete Story: How This Model Was Built, Why It's Special, and How It Works ---

### Evidence-Vector-Informed Embedding (EVIE-8B Flagship Teacher) <p align="center"> <a href="#-comprehensive-vidore-leaderboard-comparison"><img src="https://img.

### Evidence-Vector-Informed Embedding (EVIE) <p align="center"> <a href="#-comprehensive-vidore-leaderboard-comparison"><img src="https://img.

Simulated weaving

【LLMs千面郎君】AIGC 求职面试指南项目涵盖了 **📚 AIGC 求职面经**、**提示词工程**、**大模型面试题专题**、**计算机视觉与感知算法专题**、**深度学习基础与框架专题**、**手撕项目代码专题**、**优异开源资源推荐专题** 等 AIGC 求职你所需要知道的一切~

**4-bit dynamic quant.

A spaceship game played with real Hyprland tiled windows

Awesome AI for mass spectrometry papers.

Projeto da Imersão Engenharia de Dados — setembro/2026. Análise de dados da aviação brasileira com dados da ANAC, Databricks, Python e SQL.

应县木塔 · 一木千年：六幕滚动三维科普网站与可编辑模型

Codex-style Computer Use for DeepSeek Harness: the official window2 13-method desktop surface plus Chromium automation, a visible overlay with Esc interrupt, per-app approvals and bundled skills. Windows, MIT, no Codex required.

InSAR / SAR 开源软件精选索引与源码阅读笔记

Practice the AI-assisted coding interview format with your own coding agent: ten problems with planted bugs, stubs, timed tests, and an interviewer prompt

水仙的Threads 爆款雷达：实时监控 Threads 上关键词相关的公开帖子，按互动量与上涨速度排出榜单；同时按账号聚合同行，看谁在持续产出爆款。零第三方依赖（纯 Python 标准库），含关注名单、CSV 导出、互动增长曲线、通道自检与自动熔断。Real-time Threads viral content radar — monitor keyword trends, rank posts by engagement and growth velocity, and track creators who consistently produce viral content.

compress ampr (or no ampr) games into LZ4

东海赤潮风险研判技能：历史查询、真实 LightGBM 下一月预测、可复现程序与评委演示

知乎 AI 蓝宝书 · 三册文稿（AI Infra / RL / 具身智能）

Qwen3.8-27B EXL3 (4.00 bpw) + DFlash2 speculative decoding for ExLlamaV3, validated at 262k context on a 24 GB RTX 3090

Reverse-engineered Python API for prism.openai.com

Live Benchmark for long Horizon tool using Agentic Systems

自托管多渠道开发协作 Agent 系统

Long MiniMax H3 videos clip by clip with continuity: motion/identity transfer and style transfer/retexture, resumable, one clip in VRAM at a time

微信群总结报告 Skill：读取 Windows 微信本地记录，生成有出处的 PNG、HTML 和 Markdown 摘要。

Voice control for Omarchy, powered by the Gemini Live API. Talk naturally; it acts while you're still speaking.

**8-bit dynamic quant.

Long-term design specification skill for responsive web and mobile products.

Low-Latency 720p 60 FPS PC Game Streaming on a Stock, Unmodded Nintendo Switch Lite

A multiplayer browser game implemented with Python, where an AI acts as the dungeon master and players type out their actions in their own words!

UniQuery4R: Unified 4D Scene Reconstruction from a Single Query

A multimodal Web-Agent built on Qwen2.5-VL-3B, trained with Protocol-SFT and GRPO to autonomously use real visual and text Web search for knowledge-intensive visual question answering.

Record every stroke from a Concept2 PM5 over Bluetooth, force curve included, with a live dashboard. Runs in the browser, nothing to install.

Devin CLI equal-length system-prompt patcher (chisel / devin.exe)

Deepseek v4.1-Flash on a 24GB Macbook pro

Autonomous memecoin trading desk powered by 10 GPT-6 ASTRA agents and a digital fruit-fly brain

Memory Wiki dashboard plugin for Hermes Agent: browsable session-history wiki + persistent-memory audit panel

Reproducible LLM pretraining and audit-replay harness behind the OPEN model collection

Claude Code skill + hook that keeps chat corrections out of commit messages, PR bodies, comments, test names and filenames

Hermes ↔ Grok Bot bridge: SSH relay and native API (:8642) methods

| Models | Activated Params.

**One model for text-to-video, image-to-video, body- and hand-pose control, omni-reference generation, video editing, frame-by-frame repair and streaming long video — generated with four denoiser...

**Quickest burst chat speeds.

Official repository for "How You Ask Matters! Adaptive RAG Robustness to Query Variations".

Qwen3.8-Flash-Next (W4A16) on 2x CMP 170HX: 110 tok/s decode, 7K prefill, 800 tok/s @ 23 concurrent - runs on 38 GB home RAM, no pinned memory

Turn robot footage into a fly-courier delivery HUD

这是一个文案优化的agent，实现了 web-search-tool、工作流、记忆、多轮对话、skill 等agent必备能力。

投镜 · Personal Investment Twin — an AI investment decision coach: rebuilds your investment episodes from your own trading history, attributes P&L with deterministic computation, and reminds you before you repeat past mistakes. No stock picks, no prediction, no auto-trading.

Your app is guilty until proven secure: an agent skill whose /hack-me breaks into your own running app, proves each bug with a real request, patches it, and re-verifies. Claude Code · Codex · Cursor.

Uncensored Qwen3.8-27B Ultimate - BF16 master + NVFP4-MIXED deploy. QuickStarts for DGX Spark (one Dynamic DFlash lattice seat @ util 0.80), dual-Spark TP=2, RTX 5090 / PRO 6000. Serve with aeon-vllm-ultimate / aeon-vllm-ultimate-rtx.

这是一个专供华为杯数学建模的Skill，旨在打造比赛全流程一站式完工的工具包，初版于2026年9月15号上线

Humanizer Skill 3.0: Make AI Writing Sound 100% Human Locally! - Local AI prose humanizer skill for Ollama, Claude Code, and Skills CLI

Give any song a one-of-a-kind music-reactive Three.js voxel planet.

Shrinkflation detector for Codex

Open-source AI security tool to scan and fix vulnerabilities in your vibe-coded projects before you ship. AXguard by AwareXone.

Manage Cheat Engine tables and live table controls from SteamOS Game Mode. A Decky Loader plugin.

| Models | Activated Params.

**Quickest burst chat speeds.

⬢ AutoDorking v4 ELEGANT — Black Transparent × Purple Neon — 2026 Engine by Gustiakmal | Next.js 15 / AI-LLM / Supabase / R2 / tRPC | 14 kategori 60 dorks | Recon pasif only

淘宝客 MCP · 纯 Python 直连官方 API · 不经任何第三方

Portable, safe, versioned configuration packages for Frappe Framework.

Enter a company name. AI builds a traceable, computable, verifiable operating model — Generic Company Operating Twin (A-share + US)

One-click glTF 2.0 / GLB importer for Autodesk 3ds Max 2024-2026. PBR materials, textures, hierarchy, skin and animation. Created by Fady Derias.

国科大 Windows 桌面助手：托盘后台运行、轻新课堂、讲座预约、选课规划与自动化、国科大在线视频/PDF；模块化集成与详细使用说明。

Requests energy for when the dependency is the problem.

Turn a closet walkthrough video into a searchable wardrobe catalogue, with a local MCP server and agent skill

Research preview: mine coding-agent feedback into reviewable instruction changes.

DenseK3-4B 是一个把 Kimi K3 的 KDA、MLA、AttnRes 和 SiTU 等核心架构下沉到 4B Dense scale 的研究项目；我们没有从零重新预训练，而是以 Qwen3.5-4B-Base 为 donor，通过分阶段白盒架构迁移、Joint Recovery 和多教师蒸馏，把一个成熟 pretrained model 改造成新的 Dense K3-style architecture，并进一步研究它的能力保留和长上下文状态效率

Qwen3.8-27B inference, CUDA/Triton kernels, reference baseline, and benchmark results.

Self-contained Docker Pro Lab for pentest practice: boolean blind SQLi + hashcat cracking, web path traversal, command injection, internal network pivoting and SUID PATH-hijack privesc — with live flag validation and completion certificate. No VPN needed.

A local high-performance speech transcription WebUI based on transcribe.cpp + Moss-Transcribe-Diarize.

Turn an Elgato Stream Deck into a live status panel for a Proxmox VE host, running headless in its own unprivileged LXC.

Speculative decoding draft-model verification engine calculating acceptance probabilities

Mac용 Gemini 3.5 Transcribe 기반 실시간 AI 음성 전사 & 스마트 노트 데스크톱 앱

Acoustic vocal emotion, prosodic valence, and customer frustration analyzer

One sentence -> a rigged, animated voxel creature. Claude Code skill + tools: fal Seedream/Hunyuan-3D -> voxelizer -> headless Blender rig with 13 clips -> glTF

Sentence-level RAG factual grounding and hallucination detection engine

Cross-category e-commerce product bundle pricing optimizer maximizing AOV

Cost-aware Codex skill: Astra plans and verifies, Luna executes.

[EXPERIMENTAL] Full Linux environment on Android phones — native boot, no chroot, Termux, PRoot, containers, or Android compatibility layers.

Speculative decoding draft-model verification engine calculating acceptance probabilities

A persistent LLM-maintained knowledge system for technical manuals, SOPs, and operational documentation.

Cross-category e-commerce product bundle pricing optimizer maximizing AOV

Autonomous refund and dispute arbitration engine validating carrier proofs and policies

Making a Fruit fly synapse play DOOM

Ultimate assistant for R.E.P.O. Features: God Mode, Infinite Health/Stamina/Energy, Infinite Scrap, No Weapon Cooldown, Customizable Stats, Money Hack, Freeze Enemies, Blind AI, Round Director + Credits, and more. 18+ functions.

My Pi coding agent setup: skills, Exa search, model preferences, and an interactive installer.

Sentence-level RAG factual grounding and hallucination detection engine

Autonomous refund and dispute arbitration engine validating carrier proofs and policies

Acoustic vocal emotion, prosodic valence, and customer frustration analyzer

Crowd-v1 is the first official model architecture released for **CrowdGPT**, a community-driven distributed AI project.

**Flat 4-bit quantization.

> Runs with a plain `pip install "turboquant-mlx-full>=0.

| Models | Activated Params.

llama.

- Clean Shan-only vocabulary (75 characters) - Expanded embedding ready for fine-tuning - Base: F5TTS_v1_Base

Ronin is a Krea 2 Turbo checkpoint created through iterative LoRA training and merging into a selected community base model.

Blog | API | AWS | Azure | GCP | Arxiv `jina-ocr-v1` is an end-to-end document parsing model designed for high-quality OCR at an _efficient serving point_.

**QAT-optimized weights preserved at Q4_0, overhead tensors quantized to NVFP4.**

**LuminaV** is a master-free, memory-efficient adaptive optimizer engineered specifically for deep learning workloads running directly in low precision (`FP16` / `BF16`) without maintaining redundant...