Havuz

MODA: open fashion retrieval benchmark and models by Hopit AI. MODA (203M, open source), MODA Pro Lite (213M, open weights), MODA Pro (hoste

A token-efficient, provider-agnostic vision layer for reasoning models via MCP.

Design-fidelity QA for DeepSeek Harness: lend any text-only model an eye, then judge whether the implementation matches the mock. Ships the

Official code for InfoVis: one 9B model for slide image generation and editing, with the VisSlideBench benchmark and 15M+ training records.

A catalog of every MLP neuron and attention head of Qwen3-8B and Gemma 4 12B, read from the weights through two Jacobian lenses (R and S), f

[MM' 2026] Light-MER for efficient multimodal emotion recognition with sub-1B MLLMs.

Benchmarking data agents for verifiable analytics over heterogeneous workspaces.

Free unlimited Gemini API for any OpenAI-compatible app.

A multimodal RAG dashboard and interactive 3D Knowledge Graph. Process documents locally with Ollama or via Cloud APIs, powered by LightRAG,

Uni-ViGU: Towards Unified Video Generation and Understanding via A Diffusion-Based Video Generator

Cooking AGI / AGI 大锅烩 · Bilingual (English / 中文) ML & LLM notes with worked examples, step-by-step figures and coding practice. Free to read

OpenCode vision support: automatically saves pasted images and guides non-vision models to recognize them via tool calling. Supports single

Give your agent eyes: any video link in, one typed JSON out. A Claude skill + API.

Build world models from scratch — from perception and latent dynamics to planning, JEPA, VLA, and 3D/4D intelligence.

Local NotebookLM for Claude Code via Google Antigravity (agy / Gemini 3.x): /agy:notebook turns a folder of documents into per-doc summaries

A unified Model Context Protocol server for MiniMax CLI (mmx)

OpenAI-compatible reverse proxy for free multimodal AI APIs (chat / images / videos / audio / 3d)

A beautiful Gemini-powered chat interface with multimodal support

Search knowledge by what documents mean and how they look — not one or the other.

Multimodal AI studio powered by Qwen3.6-35B-A3B. End-to-end web app exposing visual reasoning, image captioning, and document understanding

让非多模态模型（DeepSeek 等）把图片识别委托给任意 OpenAI 兼容的多模态模型。零厂商硬编码，主模型支持视觉时自动隐藏自己。

基于 Google DeepMind 的 EmbeddingGemma 2 的本地多模态文件语义搜索器：用一句话描述文件的内容，在你选定的文件夹中找到文档、照片、截图、录音和视频，也可以以图搜图。支持 Windows、Android。

A zero-dependency Python tool to resolve relative Markdown links and inline images as Base64 strings for RAG and Multimodal LLMs.

Use your X Premium or SuperGrok subscription in pi.

Android 16 fork. AI as a platform primitive. Twelve capabilities, one shared runtime, every app. OEM-pluggable. Apache 2.0.

Beyond Time Shifts (ECCV 2026): a reference-free audio-visual synchronization evaluator built on Qwen2.5-Omni-3B, with SynthSync data, R-GRP

Introducing Muse Glimmer: open-weight 30B agentic multimodal model that runs on your device (Meta). Interactive local agent lab + guide. Apa

NAS OS

Open-source local-first computer-use agent for Linux, Windows, and macOS, powered by local VLMs or compatible APIs.

🐟 大肥鱼：QQ AI 伙伴 + 小红书浏览与草稿测试版 | macOS 原生 / Windows WSL2 Web · 自定义人格 · 独立记忆 · 表情与识图生图 · Photos 图库 | QQ 可用，微信测试中，持续迭代

An open-source, self-hosted social AI agent across QQ, Telegram, Feishu/Lark and personal clients, with persistent identity, long-term memor

小红书图文笔记 Agent Skill，支持看图理解、主动提问、智能选题、实时热梗搜索、标题正文、标签评论钩子和风险检查。

Black-box LLM distillation for any verifiable task, with text or image inputs: teacher API -> rejection sampling -> SFT -> eval -> GRPO

Generate editable scientific SVG figures from method text with local SAM3 and dual-provider routing.

🔥🔥 Accelerating decision inference with exact state forking

macOS computer-use agent in the notch. Long-press, talk, Claude drives the mouse.

Local-first visual media plugin for DeepSeek Harness: local video preprocessing, native vision, copyable image history, and video playback.

甲作 Atelier：手机端 AI 美甲推荐与虚拟试戴；仓库公开其中的轻量评测工作台。

Evolving what multimodal agents see and how they read it. Framework, ten reusable task skills, and paper resources.

An agent skill that reads a photo the way Sherlock Holmes reads a stranger (福尔摩斯): observe first, deduce second, every conclusion cites its

Unified native AI runtime for Java 17+, Java 22+ (Panama FFM), Android & C++. Run GGUF LLMs, Vision, Video, Whisper ASR, TTS & Embeddings in

Evidence-aware Bilibili video research MCP for language, vision, and multimodal analysis.

One embedding space for text, image, video, audio, thermal, motion, and touch. Open weights, self-hostable.

Open-source video compiler that turns videos into structured knowledge for AI agents.

NexcoreVision OCR + CNN + VLM evidence architecture, consent-gated local ComfyUI/Qwen deployment, and Agnes media generation in one Agent Sk

A small chat model trained from scratch: tokenizer → pretraining → SFT → RL → multimodal. 0.78B params, CORE 0.248, tool calling, vLLM. Weig

Long-term multimodal memory for personal agents.

Codex Skill：让纯文本模型（DeepSeek）借助 StepFun step-3.7-flash 获得看图能力 | Give text-only Codex models (DeepSeek) image understanding via StepFun step-3

Cognitive-structured Multimodal Agent (CMA-Harness): a memory-centric agent for long-horizon multimodal understanding, generation, and editi

Capability-aware vision + paste extension for the pi coding agent. Delegates image analysis only when the active model is text-only; passes