Open-source AI 28

Open-source browser AI components for private, offline applications LocalMode UI is an open-source, MIT-licensed toolkit and shadcn-style registry for building privacy-first AI interfaces that run entirely in the browser. It provides more than 100 copy-owned React components and 36 installable blocks for chat, RAG, vector search, vision, audio, agents, privacy, and developer tools. Models run locally through WebGPU or a WebAssembly fallback, so applications can work without servers or API keys and keep user data on the device.

Open-source personal AI agent workstation with private memory and multi-channel access QwenPaw, formerly known as CoPaw, is an open-source personal AI agent workstation developed by Alibaba's AgentScope team. It uses a local-first architecture so users retain ownership of their data, while supporting both local and cloud deployment. QwenPaw includes the ReMe self-evolving memory system, connects to more than 10 communication channels such as DingTalk, Feishu, QQ, Discord, Telegram, and WeChat, and allows one instance to host multiple agents with independent memories. Its Skills and MCP extensions support tool use, agent collaboration, and memory management.

Open-source AI tool for real-time face swaps and avatar creation for VTubers and streamers. DeepLiveCam is an open-source, privacy-first AI tool designed for VTubers, creators, and streamers. It offers real-time face swaps and avatar creation without requiring coding knowledge. The application allows users to morph their face into any person in real-time for streaming, live shows, and video face swaps. It works offline, doesn't collect data, and is fully local, ensuring privacy and uninterrupted use.

All-in-one real-time AI video and image generation platform. LTX Studio is an all-in-one AI video generation platform powered by Lightricks' LTX 2.3 model, which is the industry's first real-time AI video generator. The platform features a multi-model open ecosystem that integrates top-tier AI models including LTX, Veo, Kling, Seedance, Nano Banana, GPT Image, and Grok into a single workspace. It supports multimodal creative creation, allowing users to perform text-to-image, image-to-image, text-to-video, image-to-video, audio-to-video, and video-to-video transformations. Built on open-source principles, it can be deployed on private servers and embedded directly into production pipelines for individual creators and enterprise teams alike, while offering advanced capabilities like character consistency, scene modifications, lip-syncing, and audio-visual synchronization.

Open-source AI model to convert PDFs into customizable audio outputs. PDF2Audio is an open-source AI model that transforms PDFs into flexible and customizable audio outputs. It allows users to create podcasts, lectures, and summaries from PDF documents. The tool uses OpenAI GPT models for text generation and text-to-speech conversion, offering features like multiple PDF uploads, customizable instruction templates, model customization, different speaker voices, and introductory instructions.

Open-source AI unifying workflow, connecting knowledge, and automating tasks. Amurex is an open-source AI that unifies your workflow by connecting and organizing knowledge across your existing tooling. It allows you to search, automate, and retrieve insights effortlessly without switching contexts. It is fully self-hostable for complete data control and acts as an invisible companion for your work and life.

Multimodal AI for instruction-based image editing and generation. DreamOmni2 is an open-source multimodal AI model designed for instruction-based image editing and generation. It allows users to transform images by referencing abstract attributes like texture, material, and style, or by manipulating concrete objects. DreamOmni2 is highlighted for its superior identity consistency and editing precision compared to commercial AI models. It unifies multimodal instruction-based editing and generation, supporting both text and image inputs for precise control over transformations.

Private LLM is an offline AI chatbot for Apple devices, ensuring privacy and security. Private LLM is a local AI chatbot for iOS and macOS that works offline, keeping your information completely on-device, safe and private. It doesn't need the internet to work, so your data never leaves your device. It stays just with you. With no subscription fees, you pay once and use it on all your Apple devices. It's designed for everyone, with easy-to-use features for generating text, helping with language, and a whole lot more.

Open-source AI video generator with chunk-wise prompting for consistent, controllable video creation. MAGI-1 is an open-source, autoregressive AI model for video generation developed by sand-ai. It offers high temporal consistency and controllable generation through chunk-wise prompting. MAGI-1 generates videos by predicting sequences of video chunks, ensuring smooth and consistent output with fine-grained control. It supports features like image-to-video conversion and physics simulation.

Open-source desktop AI workforce for automating browser, app, and file-based tasks. Eigent AI is an open-source desktop multi-agent workforce that connects to users' context and automates real work across browsers, desktop applications, files, and other environments. Users can work with a single agent or coordinate specialized agents for research, writing, data analysis, browser automation, file processing, and multi-step workflows. Eigent supports local-first execution, scheduled automations, context connectors, multiple AI models, and self-hosting with personal API keys or local models.

Open-source platform for designing, deploying, and monitoring AI workflows. Draft'n run is an open-source AI agent platform that empowers teams to design, deploy, and monitor AI-powered features and production-ready AI workflows faster than ever. It transforms complex AI integration into an intuitive, visual process with full transparency, control, and scalability. The platform enables developers, product teams, and software and AI agencies to build production-ready AI workflows without code, including AI chatbots, automated workflows, and agentic AI, with built-in DevOps, observability, and governance. It features a drag-and-drop Studio for building workflows, an interactive Sandbox for testing, comprehensive Observability for monitoring, and REST APIs for integration.

AI image generator and editor for text, references, and bilingual posters LLaDA-Image is an independent browser-based AI image generator and editor powered by the open-source LLaDA-Image model family. It supports text-to-image generation, reference-based instruction editing, bilingual English-Chinese poster creation, and faster drafts through the LLaDA-Image Turbo model. Users can generate images from prompts, upload references, describe visual changes, and access Base or Turbo workflows online.

Platform for building AI infrastructure, natural language QA, and community collaboration. Incribo is a platform that allows users to browse and build their own AI infrastructure from a catalog of open-source AI models. It facilitates community engagement, team collaboration in content creation, and provides natural language QA for voice agents, offering downloadable, high-quality audio tests in 60 seconds.

AI-driven code intelligence models for software engineering and coding tasks. Kat Dev is an AI-driven code intelligence offering developed by Kuaishou's Kwaipilot team. It is a family of large language models specialized for software engineering and coding tasks, designed to act like AI engineers. Kat Dev models are capable of understanding complex requirements and generating high-quality code, achieving high performance on benchmarks like SWE-Bench. The models are high-parameter LLMs (32B or 72B) trained specifically for coding with state-of-the-art architectures and extensive reinforcement learning, including multi-stage agentic training with tool use, multi-turn dialogue, and coding knowledge.

Open-source AI model based on Llama 70B, surpassing GPT-4 with self-correction. Reflection 70B is an advanced open-source AI language model built on the Llama 70B architecture. It distinguishes itself by surpassing GPT-4 through innovative self-correction techniques. It excels in handling complex tasks with high accuracy and supports a wide range of applications. It is designed to engage in human-like conversation, assist with various tasks, and provide information across a wide range of topics.

Platform for creating, discovering, and chatting with virtual AI characters. Charstar is a platform where users can create, discover, and chat with virtual AI characters. It leverages the latest advancements in open-source AI to provide an unfiltered virtual character experience. Users can explore various characters, including NSFW options, and engage in conversations with them.

OmniGen AI is a free online AI text-to-image generator with editing capabilities. OmniGen is an advanced text-to-image AI model that combines text-based image creation, editing, and visual-conditional generation in a unified framework. OmniGen seamlessly transforms prompts into high-quality, visually rich outputs with no extra tools needed. It is a free online text to image generator, enabling users to create consistent images with simple prompts.

Launchpad for AI startups using tokenization and DAOs to democratize AI development. Bitlauncher is a pioneering launchpad designed to redefine innovation for open-source AI startups. It addresses funding and organizational challenges through tokenization and DAOs, aiming to democratize and decentralize AI development and ownership. The platform facilitates fundraising, token sales, and community governance for AI and Web3 projects.

Deepseek's unified multimodal AI model for understanding and generating images and text. Janus Pro AI is a unified multimodal understanding and generation model developed by Deepseek. It is an advanced version of Janus, incorporating an optimized training strategy, expanded training data, and scaling to a larger model size. Janus Pro AI excels in both multimodal understanding and text-to-image instruction-following capabilities, while also enhancing the stability of text-to-image generation. It supports bidirectional image understanding and generation via an autoregressive framework with a unified Transformer architecture.

A hub for open-source AI workflows with pre-configured templates and resources. Mozilla.ai’s Blueprint Hub provides tools to explore, collaborate, and build with open-source local models using pre-configured templates for common applications. It is a developer-first hub for open-source AI workflows, offering blueprints, building blocks, tools, datasets, and models to help users quickly start building.

Platform for creating and integrating AI agents and personal AI assistants. Mindverse AI offers MindOS, a platform to create AI agents and integrate them into daily life or business. Me.bot, a personal AI assistant, integrates with your mind and the world, managing schedules, organizing thoughts, sparking creativity, and preserving memories. Mindverse AI focuses on creating AI as a digital self, offering more than just AGI, but a networked extension through Second Me and Me.bot.

Shared GPU infrastructure and community for running open AI models through an OpenAI-compatible API. NaN is a paid community and shared GPU inference platform for builders who want to run open AI models without managing their own infrastructure. It provides an OpenAI-compatible API, shared dedicated GPUs, EU-based processing, zero prompt and response logging, and access to language, embedding, reranking, text-to-speech, and speech-to-text models. Members can use the shared inference cluster, participate in a private Discord community, attend events, and vote on future models. The platform also offers a premium GLM 5.2 tier with published token allowances.

A free website to chat with Meta's Llama 4 Maverick AI model. Chat With Llama is a free website that allows users to talk with Meta's latest AI models. Currently Meta's latest model is the 17 billion parameter Llama 4 Maverick. This model is comparable to GPT-4o and Gemini 2.0. Using Chat With Llama, you are able to test the model completly for free. This gives developers a look as to what the model is capable of. Llama 4 Maverick is a new 17 billion parameter model developed by Meta. Llama 4 was released on April 5, 2025 and is the most powerful chat model yet. It is able to beat both Gpt-4o and Gemini 2.0. Likewise, as with all previous models Llama 4 is open source so users are able to train their own version and use it commercially.

Uncensored AI app store with private and secure access to various AI models. FreedomGPT is an uncensored AI app store providing access to various AI models. It emphasizes privacy, security, and accessibility across different platforms. The platform includes the Liberty model, which answers questions without censorship. Users can access and vote on AI responses to personalize the system and select the best AI for their needs.