Top Open Source AI Dev Influencers in March 2025

A lightweight, powerful framework for multi-agent workflows

AI SYSTEMS TRANSPARENCY FOR ALL! - LEAKED SYSTEM PROMPTS FOR CHATGPT, GEMINI, GROK, CLAUDE, PERPLEXITY, CURSOR, WINDSURF, DEVIN, REPLIT, AND MORE!

Open Source AI coding assistant for planning, building, and fixing code. We're a superset of Roo, Cline, and our own features. Follow us: kilocode.ai/social

Agent Reinforcement Trainer: train multi-step agents for real-world tasks using GRPO. Give your agents on-the-job training. Reinforcement learning for Qwen2.5, Qwen3, Llama, and more!

A community-driven AI automation framework that builds upon the incredible work of the open source community. Our goal is to combine language models with specialized tools for tasks like web search, crawling, and Python code execution, while giving back to the community that made this possible.

SpatialLM: Training Large Language Models for Structured Indoor Modeling

ClaraVerse is a privacy-first, fully local AI workspace featuring a Local LLM chat powered by LLama.cpp, along with support for any provider, tool calling, agent building, Stable Diffusion, and n8n-style automation. It requires no backend or API keys—just your stack and machine.

Skywork-R1V is an advanced multimodal AI model series developed by Skywork AI (Kunlun Inc.), specializing in vision-language reasoning.

Data transformation framework for AI. Ultra performant, with incremental processing.

Run LLMs with MLX

这是一个基于 FastAPI 构建的 Gemini API 代理

RikkaHub is a Android APP that supports for multiple LLM providers.

🌐 WebThinker: Empowering Large Reasoning Models with Deep Research Capability

Awesome RL Reasoning Recipes ("Triple R")

verl-agent is an extension of veRL, designed for training LLM/VLM agents via RL. verl-agent is also the official code for paper "Group-in-Group Policy Optimization for LLM Agent Training"

📎 Clippy, now with some AI

An open-source tool for general prompt optimization.

A reimplementation of Stable Diffusion 3.5 in pure PyTorch

High Performance LLM Reverse Proxy

Latest Advances on Long Chain-of-Thought Reasoning

An open-source solution for full parameter fine-tuning of DeepSeek-V3/R1 671B, including complete code and scripts from training to inference, as well as some practical experiences and conclusions. (DeepSeek-V3/R1 满血版 671B 全参数微调的开源解决方案，包含从训练到推理的完整代码和脚本，以及实践中积累一些经验和结论。)

[Fully open] [Encoder-free MLLM] Vision as LoRA

An MCP server for interacting with Sentry via LLMs.

Reproduction of DeepSeek-R1

Stay ahead of AI trends with automated Reddit insights! 🚀 This tool scans AI-related Reddit communities in English & Chinese, using Reddit Official API, DeepSeek R1 host by Groq to analyze posts, summarize key discussions, and track trends. Daily rankings highlight hot topics—catch emerging trends before they go mainstream! (Updated every 6 AM CDT

[CVPR 2025, Spotlight] SimLingo (CarLLava): Vision-Only Closed-Loop Autonomous Driving with Language-Action Alignment

The MCP manager. Easily manage your MCP servers with enhanced security and customizability

Ollama负载均衡服务器 | 一款高性能、易配置的开源负载均衡服务器，优化Ollama负载。它能够帮助您提高应用程序的可用性和响应速度，同时确保系统资源的有效利用。

Bella OpenAPI是一个提供了丰富的AI调用能力的API网关，可类比openrouter，与之不同的是除了提供聊天补全(chat-completion)能力外，还提供了文本向量化(text-embedding)、语音识别(ASR)、语音合成(TTS)、文生图、图生图等多种AI能力，同时集成了计费、限流和资源管理功能。且集成的所有能力都经过了大规模生产环境的验证。

LLMVoX: Autoregressive Streaming Text-to-Speech Model for Any LLM

DeepSeek-V3/R1 inference performance simulator

Code for the paper "Vamba: Understanding Hour-Long Videos with Hybrid Mamba-Transformers" [ICCV 2025]

Muice-Chatbot 的 Nonebot2 实现 | 支持调用 QWQ-32B & DeepSeek-R1 等主流大模型 | 支持 Function Call 和内置 MCP Host 实现

A Go-based utility that processes input through multiple AI models concurrently (OpenAI, Claude, and Gemini) and provides a summarized comparison of their responses

Summary of the Specs of Commonly Used GPUs for Training and Inference of LLM

A list of open-source AI projects you can use to generate income easily.

FakeVLM: Advancing Synthetic Image Detection through Explainable Multimodal Models and Fine-Grained Artifact Analysis

Whisper STT + Orpheus TTS + Gemma 3 using LM Studio to create a virtual assistant.

A blog theme for Halo

pdfLLM is a completely open source, proof of concept RAG app.

这是一个高性能的LLM推理API，且自带UI界面，它将DeepSeek R1 的思维推理信息和Anthropic的 Claude 系列模型相集成。

A curated list of awesome works in Routing LLMs paradigm (👉 Welcome to submit your contributions to this code repository)

SORT3D, an LLM-based object-centric grounding and indoor navigation system employing a spatial reasoning toolbox and state of the art 2D VLMs for perception.

VulnWatchdog 是一个自动化的漏洞监控和分析工具。它可以监控 GitHub 上的 CVE 相关仓库,获取漏洞信息和 POC 代码,并使用 GPT 进行智能分析,生成详细的分析报告。

Hub for researchers exploring VLMs and Multimodal Learning:)

This compendium reviews significant published research contributions and industrial engineering practices in leveraging Generative AI and LLMs for developing search, recommender, personalization, and question-answering systems. It aims to cover the entire spectrum of research and practices

Awesome Reasoning in MLLMs: Papers and Projects about learning to reason with MLLMs, including Chain-of-Thought (CoT), OpenAl o1, and DeepSeek-R1

Complex Multimodal Reasoning with Reinforced LLMs/VLMs/MLLMs. Motivated by the DeepSeek-R1-Zero

mod-ollama-chat is an AzerothCore module that enhances the Player Bots module by integrating external language model (LLM) support via the Ollama API.

Public Goods Game (PGG) Benchmark: Contribute & Punish is a multi-agent benchmark that tests cooperative and self-interested strategies among Large Language Models (LLMs) in a resource-sharing economic scenario. Our experiment extends the classic PGG with a punishment phase, allowing players to penalize free-riders or retaliate against others.

Delightful AI coding mentor | Open-source, local, hackable | 🌟 Star if it sparks delight!

Intelligent Malware that takes screenshots for entire monitors and exfiltrate them through Trusted Channel Slack to the C2 server that's using GPT-4 Vision to analyze them and construct daily activity — frame by frame

A multimodal agent that can interact with its own PC in a multimodal manner.

Run AI coding agents like Cursor, Windsurf, and Claude Code via code and let them implement features all the way to a pull request

Official implementation of LangCoop: Collaborative Driving with Natural Language

Dremio MCP server

simpleR1: A Simple Framework for Training R1-like Models

Encourage Medical LLM to engage in deep thinking similar to DeepSeek-R1.

Powerful tool designed to generate SQL queries from natural language (NL2SQL) using Microsoft’s Semantic Kernel framework. This project aims to bridge the gap between human-readable queries and SQL, enabling easy and efficient database interactions with AI-driven language models.

Super simple python connectors for llama.cpp, including vision models (Gemma 3, Qwen2-VL). Compile llama.cpp and run!

基于 🦜️🔗 LangChain 与 DeepSeek R1 大语言模型的本地知识库问答 serve 端。 本项目是本地知识库问答应用的 serve 后端。目前实现基本的 RAG 功能。 使用 FastAPI + Uvicorn + SQLModel + SQLite 框架，向量数据库使用 Chroma 。vue 前端服务请详看README

a training-free approach to accelerate ViTs and VLMs by pruning redundant tokens based on similarity

This application is a simple but powerful web tool for translating subtitle files in .srt format. To translate files, use the LM Studio AI model running on your local machine or of the online models of ChatGPT-4o, ChatGPT-4o mini and some openrouter models with API keys.

Multi-modal agentic framework for surgical procedures

ChinaTravel: A Real-World Benchmark for Language Agents in Chinese Travel Planning

MCP server tailored to connecting web crawler data and archives

Build a powerful Deep Research AI agent like Gemini or ChatGPT. Using Next.js, Vercel AI SDK, and Exa Search API, An intelligent system that generates follow-up questions, crafts optimal search queries, and compiles comprehensive research reports.

Your personal deep research ai agent, a free & open source alternative to open-ai deep research

Llama post-training examples with torchtune

OpenAI GPT-4o Mini TTS – Home Assistant Integration

A benchmark that focuses on the sampling dilemma in long-video tasks. Through well-designed tasks, it evaluates the sampling efficiency of long-video VLMs. (ICCV2025)

AgentMake AI: a kit for developing agentic AI applications that support 16 AI backends and and work with 7 agentic components, such as tools and agents. (Developer: Eliran Wong) Supported backends: anthropic, azure, azure_any, cohere, custom, deepseek, genai, github, github_any, googleai, groq, llamacpp, mistral, ollama, openai, vertexai, xai

Speech Resynthesis and Language Modeling Using Flow Matching and Llama

Enhanced version of "Sequential Thinking" MCP

A complete cross-modal RAG system for end-to-end speech-to-speech large models, including ASR-based Retrieval and E2E Retrieval.

Convert LLaMA3.1-8B to DeepSeek R1 MLA & MoE (raw)

GeoGuessr benchmark for language models

A MCP server that provides text-to-image generation capabilities using Stable Diffusion WebUI API (ForgeUI/AUTOMATIC-1111)

KiCad Library Manager | Easily manage global libraries with github

This template provides a streamlined foundation for building Model Context Protocol (MCP) servers in Python. It's designed to make AI-assisted development of MCP tools easier and more efficient.

Learning and research after DeepSeek-R1, around test-time computing, resurgence of RL, and new LLM learning/application paradigms.

MLX.zig: Phi-4, Llama 3.2, and Whisper in Zig

Reproduction of the complete process of DeepSeek-R1 on small-scale models, including Pre-training, SFT, and RL.

DiffuGen is a powerful yet user-friendly interface for local\edge image generation. Built on the Model Control Protocol (MCP), it provides a seamless way to interact with various Stable Diffusion models including Flux, SDXL, SD3, and SD1.5. Diffugen also features an OpenAPI Server for API usage and designed to support OpenWebUI OpenAPI Tool use.

Open-Source Language Model Database for comparison

An interface that features barely zero external dependencies beyond the Ollama API itself, making it lightweight and portable to easily interact with your local ollama server from a PC or a even your smartphone.

AI Lawyer is an intelligent reasoning legal assistant powered by DeepSeek , Ollama RAG and LangChain, designed to streamline legal research and document analysis. By leveraging retrieval-augmented generation (RAG), it provides precise legal insights, and contract summarization. With an intuitive Streamlit-based UI, analyze legal documents.

A Unity MCP server that allows MCP clients like Claude Desktop or Cursor to perform Unity Editor actions.

A high-performance PyTorch toolkit for pre-training modern, Llama-style language models. Based on nanoGPT with significant architectural enhancements.

The opinionated high performance professional-grade AI package for Go

Safety Mirage: How Spurious Correlations Undermine VLM Safety Fine-tuning

[CVPR'25] Official implementation of the paper "Not Only Text: Exploring Compositionality of Visual Representations in Vision-Language Models".

​This repository provides code and resources for automating manufacturing feature recognition in CAD designs using vision-language models.

This provides a web UI for AI chat playground that is able to connect virtually any LLM from any platform.

JARVIS Virtual Assistant, inspired by Marvel's Iron Man, is an AI-powered tool with voice activation, system monitoring, and proactive suggestions from camera and screenshot analysis. Built with PyQt5 and OpenCV, it boosts productivity with a witty, JARVIS-like charm.

Unified API platform for free access to enterprise-grade AI models from Google, Groq, and OpenRouter. Industrial-ready integration with high-performance Models Inc. DeepSeek-R1, QwQ 32B

Proof-of-concept Aider w. local (24GB vram) QwQ+Qwen2.5-Coder using litellm-proxy / llama-swap / llama.cpp