All AI tools

Qwen2-VL

Qwen2.5-Coder

AI coding

Alibaba's open-source coding model, leading choice for self-hosted/local deployment in 2026.

Qwen3-TTS

AI audio

Qwen3-TTS is an open-source text-to-speech (TTS) model based on a 12Hz efficient tokenizer and a multi-codebook speech encoder architecture, striking a balance between compression efficiency and detail preservation, and capable of capturing paralinguistic features such as breathing, hesitation, and emotional intensity. It supports zero-shot voice cloning, requiring only 3 seconds of reference audio to replicate a speaker's timbre and style; it natively supports more than 10 languages (including English, Mandarin Chinese and dialects, Japanese, Korean, French, German, etc.) and seamlessly handles code-switching; it features context-aware prosody, natural language audio control (such as whispering, shouting, laughing, and fast speech), long-text synthesis, and real-time streaming output. Official data shows a first-token latency of 97ms. The model is released under the Apache 2.0 license. Developers can install it via pip, use the Python SDK and an OpenAI-compatible API, and deploy an OpenAI-compatible API server with a Docker image.

Qwen3-TTS

AI audio

Qwen3-TTS is an open-source text-to-speech model launched by Alibaba, supporting 10 languages, 3-second zero-shot voice cloning, prompt-driven timbre and emotion control, and featuring 97-millisecond low-latency streaming generation capability.

Qwerki

Qwerki uses AI to help you remember things and organize your life.

Qwilr

Qwilr is an interactive proposal software designed to increase deal velocity. It transforms static documents into interactive buying experiences, with built-in e-signature and payment capabilities, helping buyers make commitments instantly, thereby accelerating the sales cycle. Qwilr provides real-time buyer analytics, allowing sales teams to understand how buyers interact with proposals, and sends real-time notifications for timely follow-up. The software integrates with major CRM, sales, and accounting software such as HubSpot, Salesforce, Zoho, Pipedrive, Quickbooks, and more, automating proposal creation and post-sale management. Qwilr claims that its customer teams can achieve a win rate 44% higher than the industry average and close deals up to 40% faster. It offers proposal templates and collateral covering sales, marketing, customer success, and other scenarios. Qwilr is loved by over 5,000 customers and offers a 14-day free trial with no credit card required.

Qwizzard

AI learning

AI-powered quiz generator from uploaded files.

Qz kraft

AI learning

AI-powered quiz generator for creating customized quizzes with autograding and proctoring.

R yan anlat

AI-powered dream interpretation service offering free dream analysis.

R1-Nature

R1-Nature is an open-source project on GitHub, aiming to reproduce DeepSeek R1-like results on small models in the simplest way, and to explain through comparative experiments the most important essence of O1-like and R1-like systems—the internal reasoning process of think thinking. On small models (0.5B, 1B, 1.5B, 3B), the project compares the fine-tuning effect differences of different thinking reasoning CoT datasets, releases related datasets, fine-tuned models, LoRA weights, and GGUF models, and discusses the chain-of-thought reasoning explosion phenomenon and its possible solutions. The project believes that the procedural content of think thinking is the core of AGI/ASI, but there are still many unclear details that researchers need to pay attention to, rather than simply distillation plus RL. The content covers the software and hardware environment required for training, differences in dataset sources and label formats, fine-tuning and testing scripts, and observations about the causes and strategies of reasoning explosion. It is suitable as a reference for researchers interested in R1 reproduction, CoT, and the reasoning capabilities of small models.

R1-Onevision

R1-Onevision is a multimodal reasoning large model designed to solve complex visual reasoning tasks. It was developed by a team at Zhejiang University and can seamlessly integrate visual and textual data, excelling in fields such as mathematics, science, deep image understanding, and logical reasoning. The project converts images into formal textual representations through a cross-modal reasoning pipeline, thereby enabling precise language-based reasoning. The model is fine-tuned based on Qwen2.5-VL, and the accompanying dataset, benchmark, SFT weights, RL weights, and demo have been released.

R1-V

R1-V is a reinforcement learning training framework for vision-language models (VLMs), designed to enhance the model's super generalization ability at an extremely low cost (less than $3). It provides training datasets, training code, evaluation scripts, and supports multiple models and tasks, such as object counting, geometric reasoning, etc. The project is maintained by the community, continuously updated, and provides blogs and papers to share findings.

R1-Vision

R1-Vision is an open-source research project aimed at distilling the strong reasoning capabilities of large language models such as DeepSeek-R1 into large vision-language models, enabling them to process both text and image inputs simultaneously. Through a cold-start dataset and the Caption-Prefixing technique, the project lets a pure-text reasoning model "pretend to see" an image, thereby generating captions during its thinking process and completing reasoning. An initial version of the cold-start dataset has currently been released, using DeepSeek-R1-Distill-Qwen-32B for reasoning and GPT-4o-mini for image description and data formatting. The project is still developing an enhanced version, with plans to incorporate stronger models, increase task diversity, and improve sample quality.

R3achout

AI office

AI tool to simplify recruitment, enhance outreach, and improve candidate qualification.

Raay

AI-powered platform for creating forms and surveys with quick analysis and integrations.

rabbitOS intern

A general-purpose AI agent launched by rabbit

Raber app

AI website builders

AI-powered drag-and-drop page builder for React.

Radical human intelligence

AI-driven platform revolutionizing customer experience and mitigating PR crises.

radops

RadOps is an open-source AI-driven multi-agent platform for NetDevOps and security operations automation, capable of autonomously executing complex multi-step infrastructure operations. It uses a Supervisor-Worker architecture with strict sequential logic, features three layers of cognitive memory (working memory, temporary facts, core knowledge), and can distinguish and retain context. The platform supports quickly generating specialized agents for networking, security, and more through YAML configuration, without writing new code; it also provides human-in-the-loop approval, Markdown-defined Agent skills, and unified workflow SOPs. Its trust-but-verify mechanism uses dedicated QA audit nodes to validate actual tool output and reduce hallucinations. RadOps supports model providers such as OpenAI, Anthropic, DeepSeek, Azure, Google, Groq, Mistral, AWS Bedrock, and Ollama, and declaratively connects to vector databases such as Weaviate, Qdrant, Pinecone, Milvus, and Chroma. It implements resilient connections based on MCP and provides deep observability through OpenTelemetry. It is suitable for teams that need to automate network and security operations processes.

Raena ai

AI office

Raena AI transforms study notes into AI-powered quizzes, summaries, and learning tools.

rag-architect

The introduction is being prepared. Please check back soon.

rag-kb

Local-first agent memory and knowledge base: hybrid dense and BM25 retrieval, document and webpage ingestion and optional RAG answers via local Ollama.

rag-mcp

Minimal local RAG that vector-searches a document corpus and returns cited passages.

rag-rat

AI coding

Local repo intelligence: hybrid BM25/vector search, a tree-sitter code graph, queryable git and GitHub history, and persistent source-anchored memories.

Ragchat

AI office

Chat with your documents using AI for instant, accurate answers and seamless interaction.

ragmap

MapRag: RAG-focused MCP subregistry for discovering and routing to retrieval-capable MCP servers with structured constraints and explainable ranking.

Raia

AI-powered unified security platform for threat remediation and automation.

Rails guard

Invisible guardrails for Rails consoles with AI data masking and passwordless authentication.

rails-mcp

Self-hosted default-deny action registry, append-only spend-intent ledger and human sign-off audit trail for gating irreversible agent actions.

Railway

Railway offers reliable web hosting with top-tier security, seamless scaling, and easy deployments, ...

Raily

AI travel

Travel dating app for train travelers with AI, AR, and gamification features.

Rainbow Magic Fairy Name Finder

Rainbow Magic Fairy Name Finder is an independent fan-made lookup tool based on the Rainbow Magic book series that finds the corresponding fairy name by entering a name. The tool uses deterministic matching logic: after input, it normalizes and looks up an exact match in the name index; if a match is found, it displays the corresponding card and cover reference, and if no match is found, it clearly indicates no result and provides an option to re-enter, with no hash fallback or random substitute cards. The homepage is divided into three sections: How It Works, Introduction, and FAQ, with an intuitive flow that makes it easy for first-time visitors to understand and use. Result cards include the specific fairy title and indicate Orchard Series Books as the image source, emphasizing that copyright belongs to their respective owners. The tool is free to use, supports sharing, works on mobile and desktop, and is positioned as an identity generator and compact encyclopedia, suitable for readers to look up quickly and accurately, and can also be used as a reading guide and collection checklist.

Raindance AI Video Generator

AI video

This tool is an online YouTube subtitle extractor that instantly converts YouTube video links into transcripts. It can be tried for free without registration, and 3 free transcripts can be generated per day. It supports copying the full text, downloading in TXT/SRT/VTT formats, and using AI for translation, summarization, and other operations. A Chrome extension is provided for reading, searching, summarizing, and chatting directly on YouTube pages. It supports multiple languages of subtitles included with videos, including automatically generated subtitles. It is suitable for content creators, researchers, students, and other users who need to convert videos to text. It also provides API services and Python/Node.js SDKs, as well as automation integrations such as Zapier and n8n.

Rallied

Rallied is an AI automation tool for managed service providers (MSPs) that uses AI agents to handle onboarding, offboarding, and L1/L2 tickets within PSA systems. It can automate routine tasks such as password resets and script runs, and lets you write runbooks in natural language. Rallied supports platforms such as ConnectWise, HaloPSA, and Autotask. In Plan mode, it first shows the actions it intends to perform, and the paid plan is $150 per month including 2,000 credits.

Ramayan quest

AI-powered Q&A platform for the Ramayan.

Ramban ai