FLUX.1 Kontext is an image generation and editing model developed by Black Forest Labs that supports context-aware image processing. The model generates and edits images based on text and image cues, supporting object modification, style adjustments, and more.
SignGemma is the world's most powerful sign language translation AI model, developed by Google DeepMind. It focuses on translating American Sign Language (ASL) into English text, employing a multimodal training method that combines visual and textual data...
DGM (Darwin Gödel Machine) is a self-improving artificial intelligence system that enhances performance by iteratively modifying its own code. DGM selects an agent from its maintained archive of coded agents, generates new versions based on a base model,...
Circuit Tracer is an open-source tool developed by Anthropic for studying the internal workings of large language models. Circuit Tracer reveals the model's generative characteristics based on generative attribution graphs...
Congrong Big Model is a multimodal AI model launched by CloudWalk Technology. The model topped the multimodal benchmark list of the internationally authoritative evaluation platform OpenCompass with a score of 80.7, surpassing top teams such as Google and OpenAI.
DeepEyes is a multimodal deep thinking model jointly developed by the Xiaohongshu team and Xi'an Jiaotong University. Based on end-to-end reinforcement learning, it achieves "thinking with graphs" capabilities similar to OpenAI o3, without relying on supervised fine-tuning (SFT). ...
TrackVLA is a product-level end-to-end navigation model launched by Galaxy General. The model possesses pure visual environment perception, language command-driven operation, autonomous reasoning, and zero-shot generalization capabilities, enabling end-to-end processing from visual perception to action output...
TEN VAD is a high-performance, real-time speech activity detection system designed for enterprise applications. TEN VAD accurately detects speech activity in audio streams, featuring low latency, lightweight design, and high accuracy. TEN VAD is based on...
Google AI Edge Gallery is an experimental application launched by Google that allows users to experience and use machine learning (ML) and generative artificial intelligence (GenAI) models on their local devices. The application currently supports Android...
VRAG-RL is a vision-perception-driven multimodal RAG inference framework developed by Alibaba's Tongyi Big Model team. It focuses on improving the retrieval, reasoning, and understanding capabilities of visual language models (VLMs) when processing rich visual information. ...
Jaaz is an open-source AI design agent, a free local alternative to Lovart. It boasts powerful AI design capabilities, intelligently generating design hints and batch-producing images, posters, storyboards, and more.
OCode is a native AI programming assistant for mobile devices, providing developers with deep codebase intelligence and automated task execution. It seamlessly integrates with the local Ollama model, directly embedding enterprise-grade AI assistance into the development process. ...
Firesearch is an AI-driven deep research tool developed by the Mendable AI team. Based on Firecrawl's multi-source network content extraction technology, and combined with OpenAI GPT-4o's search planning and content generation capabilities, it can extract complex...
PlayDiffusion is a new audio editing model introduced by PlayAI. Based on diffusion model technology, it is specifically designed for fine-grained audio editing and restoration. The model encodes audio into discrete token sequences, masking the parts that need modification...
Auto Think is an open-source automatic thinking model developed by the KwaiCoder-AutoThink-preview team from Kuaishou's Kwaipilot platform. This model addresses the "overthinking" problem inherent in deep thinking models by conducting in-depth research and proposing a...
OmniAudio is a technology developed by the Speech Team of Alibaba Tongyi Labs to generate spatial audio (FOA) from 360° video. It aims to provide a more realistic audio experience for virtual reality and immersive entertainment. This is achieved by building a large-scale dataset...
OpenAudio S1 is a text-to-speech (TTS) model from Fish Audio, trained on over 2 million hours of audio data and supporting 13 languages. It employs a dual-autoregressive (Dual-AR) architecture and combines reinforcement learning with human feedback...
MoonCast is a zero-shot AI podcast generation project that synthesizes natural podcast-style speech from plain text sources. Trained using a long-context language model and large-scale speech data, it can generate podcast audio clips several minutes long, supporting both Chinese and English...
Gemini Fullstack LangGraph Quickstart is an open-source project launched by Google DeepMind to help developers quickly build a full-stack intelligent research assistant based on Google Gemini 2.5 and LangGraph.
DecipherIt is an AI-driven research assistant tool that simplifies and optimizes the research process using intelligent methods. The tool supports converting various topics, links, and files into AI-generated research notebooks, providing comprehensive summaries, interactive features, and more...
Qwen3 Embedding is a dedicated text representation, retrieval, and ranking model developed based on the Qwen3 basic model. It inherits the advanced architecture of Qwen3, such as Grouped Query Attention and the SwiGLU activation function, and...
Qwen3 Reranker is a text re-ranking model released by Alibaba's Tongyi Qianwen team, belonging to the Qwen3 model family. It employs a single-tower cross-encoder architecture, inputting text pairs and outputting relevance scores. The model undergoes multi-stage training...
Eleven v3 is an advanced text-to-speech model from ElevenLabs. It achieves precise control over emotion and intonation through inline audio tags, supports multi-speaker dialogues, and makes conversations more natural. The model supports over 70 languages and boasts advanced text understanding...
MiniCPM 4.0 is a large edge-side model launched by Wallfacer. The model comes in two parameter scales: 8B and 0.5B. The 8B lightning sparse version employs an innovative sparse architecture, enabling efficient processing of long text tasks; the 0.5B version uses low computational resources...