Top Open Source AI Dev Influencers in March 2025

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: Large Language Model for Spatial Understanding

Pioneering Multimodal Reasoning with CoT

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。)

Run LLMs with MLX

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。

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

Stay ahead of AI trends with automated Reddit insights! 🚀 This tool scans AI-related Reddit communities in English & Chinese, using DeepSeek R1 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)

DeepSeek-V3/R1 inference performance simulator

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。

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

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

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

Llama post-training examples with torchtune

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

Speech Resynthesis and Language Modeling with Flow Matching and Llama

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

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.

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.

MLX.zig: Lightweight Zig language bindings for Apple's MLX framework, enabling efficient machine learning directly on Apple Silicon with zero external dependencies.

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

Integrates OVMS V3 into Home Assistant through HACS. Communicates via MQTT, automatically creating and parsing vehicle entities. Developed with AI assistance.

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

AM1 – A personalized web application for conducting DeepResearch with any Multi-Agent Framework and LLM, designed to run seamlessly on production servers or your local machine.

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.

AgentMake AI: a kit for developing agentic AI applications that support 14 AI backends and and work with 7 agentic components, such as tools and agents. (Developer: Eliran Wong) Supported backends: anthropic, azure, cohere, custom, deepseek, genai, github, googleai, groq, llamacpp, mistral, ollama, openai, vertexai, xai

Chat with a Codebase using DeepSeek R1, Claude, GPT, Gemini

The official code of paper Critic-V: VLM Critics Help Catch VLM Errors in Multimodal Reasoning

DeepGPT4.5是一个创新的人工智能应用，将DeepSeek R1模型的思维链能力与GPT-4.5的强大推理能力相结合。通过这种结合，我们能够获得更深入、更全面的AI推理结果，同时为用户提供思维过程的透明度。

⽊ a lightweight agentic framework for deepseek-r1

DeepSeek 中文官网、网页版入口、下载及使用教程 | DeepSeek 最全教程指南~ （2025年3月更新） 轻松使用 DeepSeek 网页版，享受 快速、稳定、不卡顿 的体验，支持 DeepSeek R1 满血版~ 本指南提供全面的 DeepSeek 使用教程指南，包括 DeepSeek 官网入口、DeepSeek 网页版、DeepSeek 下载 和 DeepSeek 平替网站，助力顺畅使用 DeepSeek~

Pure RL to post-train base models for social reasoning capabilities. Lightweight replication of DeepSeek-R1-Zero with Social IQa dataset.

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

See How Big Exactly A 128k Token Text Is

全栈式AI服务平台技术方案，支持 DeepSeek，OpenAI，Midjourney，Stable Diffusion，DALL·E，Gemini等模型，支持多端使用，自定义模型对接，对话会话隔离、云端存储、支持多设备共享数据，多模态模型支持，支持图片解析，支持联网搜索，支持AI智能体应用，支持官方微信支持、易支付等支付方式，数据统计, 丰富美观的仪表盘统计。

Simple configurable wrapper for local LLM-based test refinery over llama.cpp API

Tool to give AI generated score, to analyze with patterns how much input text is AI generated using AWS Bedrock Llama 3.1 405B

DeepSeek-R1 中文翻译

Serverless DeepSeek R1 Inference with FastAPI and Lambda SnapStart

GLM深远海船舶作业大模型应用赛——个人技术方案 "多Agent协同系统“

Stable Diffusion WebUI

Данный проект демонстрирует базовую реализацию смены лица с использованием Stable Diffusion Inpainting Pipeline. Основная идея – заменить лицо на целевом изображении так, чтобы оно выглядело фотореалистично и органично, с сохранением узнаваемых черт, аналогичных референсному лицу.

Ollama中llama.cpp的Go语言绑定

well this is one of my good project

Implementation of Group Relative Policy Optimization (GRPO) to fine-tune Open Language Models like LlaMa-3.2, Qwen2 for Math Tasks.

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

LexiSynapseAI es una aplicación de escritorio desarrollada en Python que busca facilitar la vida académica de los estudiantes, integrando funcionalidades como gestión de tareas, recordatorios, calendario, y un asistente inteligente basado en IA.

Minimal script I'm using to train DeepSeek R1 1.5B on a RTX4060 8GB

This script demonstrates how to fine-tune a distilled Llama model using Unsloth for medical question answering.

LLM-With-Web-Search 是一个基于大语言模型（LLM）的智能问答系统，集成网络搜索能力以增强对话质量。

Zsh-centric command-line interface for interacting with local Large Language Models (LLMs). Chat directly on the command line with non-contiguous command line calls.

There are several basic operations for DeepSeek-R1-Distill-Qwen-7B.

This project fine-tunes a LLaMA model on Indian legal laws using LoRA and 4-bit quantization for efficient training. It enhances legal text generation with a custom dataset, optimized preprocessing, and Trainer API.

Qwen2.5 Adapter Instruct: Fine-tuning Qwen2.5 with the Llama-Adapter method for Fact Checking

Simplify context preparation for large language models. This Streamlit application uses llama-parse and files-to-prompt to combine PDF and text files into a single, Claude XML-formatted context file for LLMs.

Karpathy's llama2.c with improvements from modded-nanogpt.

webgpu based llm chatbot, try on chrome browsers

Chatty AI transforms your terminal into a vibrant hub where history's greatest minds come alive. Chat one-on-one with Einstein about relativity, watch Shakespeare and Isaac Newton debate the human condition, or let Marie Curie and Ada Lovelace discuss women in science.

llama-server in Docker.

LLM-based Q&A on preloaded docs, raw data, Wikipedia articles and scraped web pages with knowledge graphs, analytics, charts and Streamlit interface

ChatGPT 中文版：国内直连指南（支持GPT-4、4o、o1、o3 和 DeepSeek R1，无需翻墙）【3月最新更新】全面体验 ChatGPT 中文版，无需翻墙，支持 GPT-4、4o、o1、o3 和 DeepSeek R1 及中文多功能应用！ 本项目旨在为用户提供一站式的 ChatGPT 中文版使用指南，同时整理了国内可用的 ChatGPT 镜像网站和官网使用教程，帮助您快速上手 ChatGPT，无论是个人使用还是专业需求。

ChatGPT 中文版：国内直连镜像网站汇总合集（支持 GPT-4、GPT-4o、o1、o3、 DeepSeek R1，Grok-3 无需翻墙）【2025/03/04 更新】

Создание телеграм бота с ChatGPT o1, o3-mini, DeepSeek, Claude 3.7, Command-A, MiniMax

欢迎体验基于**RAG（检索增强生成）**架构的智能问答系统！📚 这个小助手不仅能读懂你的PDF，还能像个聪明的小伙伴一样回答你的问题！✨ 我们用Ollama作大脑，通过向量检索让文档问答变得更聪明、更高效！🧠

A web-based application for generating AI-powered images using Stable Diffusion.

Use RAG to build web app to recommend trending movies based on users' moods. Tech stacks: TMBD API, Python, LangChain, GPT-4o-mini

Generate Pokémon images from text with Stable Diffusion and an interactive Gradio interface.

PickHacks - Hackathon2025 || Winner - 3rd Place Overall & Winner - Best use of Gemini API

A practical Python reference manual covering everything from data wrangling to machine learning, deep learning, and AI—complete with ready-to-use code snippets, interactive examples, and downloadable Jupyter notebooks.

一个利用LLM能力进行翻译并自动修订翻译结果的工具，可视化展示修订前后的差异。A tool that leverages LLM capabilities to translate and automatically revise translation results, visualizing the differences before and after revision.

run llama.cpp quickly and conveniently.

A Very powerful summarization tool that processes and extracts key insights from content using Llama 2, Haystack, Whisper, and Streamlit. Runs smoothly on CPU with Llama

Deepseek R1 and Qdrant-Powered Financial Doc Intelligence

本地部署DeepSeek-R1接入微信机器人，实现个性化微信助手

This project integrates Segment Anything Model (SAM) with Stable Diffusion and LaMa for mask-free image inpainting. Users can click on objects to remove, fill, or replace them with AI-generated content.

Link download👇

Maintenance of a set of tools to enhance LLM through MCP protocols.