Latest AI Resources

GLM-OCR 是智谱开源的 0.9B 轻量级专业 OCR 模型，在 OmniDocBench V1.5 以 94.6 分刷新 SOTA。兼顾“小体积”与“全场景”，扫描、手写、印章、多语混排、复杂表...

DPAI Arena (Developer Productivity AI Arena) is an open benchmarking platform created by JetBrains to measure the effectiveness of AI-assisted development tools in real-world software engineering tasks. Through a transparent evaluation stream...

Chatterbox-Turbo 是 Resemble AI 推出的开源文本到语音（TTS）模型，专为高效、低延迟的语音合成而设计。基于350M参数的精简架构，单步推理生成音频，时间延迟极低，在150...

Dexbotic is the open source Visual-Linguistic-Action (VLA) model of embodied intelligence one-stop scientific research service platform of Dexmal, which solves the problems of fragmentation and low efficiency of research in the field of embodied intelligence. Based on PyTorch, Dexbotic is a one-stop research service platform to solve the problems of fragmentation and inefficiency in the field of embodied intelligence...

MOSS-Speech is an open source speech-to-speech (Speech-to-Speech) big model by Prof. Qiu Xipeng's team at Fudan University. It breaks through the traditional speech processing, without the need for text guidance, and directly understands and generates speech, which can capture non-text elements such as intonation and emotion, making...

Jamba Reasoning 3B is a lightweight inference model open-sourced by Israeli AI startup AI21 Labs with strong performance and potential for a wide range of applications. It utilizes a hybrid SSM-Transformer architecture that combines Trans...

Kosong is a new AI Agent development framework open-sourced by Dark Side of the Moon (Moonshot AI) that provides developers with a lightweight, flexible, and highly scalable underlying support for building next-generation intelligent body applications. With an asynchronous tool orchestration engine that efficiently schedules multiple tools...

Qwen3.5是阿里巴巴通义千问团队开源的最新一代大语言模型，属于千问（Qwen）系列的升级版本。Qwen3.5突破了传统文本模型的限制，实现了真正的原生多模态理解，可直接处理文本、图像、视频等多种模...

FunctionGemma is a lightweight AI model optimized for function calls launched by Google, developed based on the Gemma 3 base model with 270 million parameters, which can convert natural language into executable API instructions in real time on cell phones, browsers and other devices. The core feature is support for local off...

10Kh RealOmni-Open是简智机器人开源的具身智能数据集，是行业内规模最大的开源具身智能数据集。数据集累计拥有超10000小时数据、100万+片段，覆盖10大场景任务、超过30项技能。数据...

SCAIL (Studio-Grade Character Animation via In-Context Learning) is a film and television grade character animation generation framework proposed by Smart Spectrum in collaboration with Prof. Liu Yongjin's group at Tsinghua University. Through...

Pyscn is an intelligent code quality analysis tool designed for Python developers to detect potential problems in code to improve maintainability. It analyzes dead code through control flow diagrams, identifies duplicate code using APTED+LSH algorithm, calculates metrics such as module coupling and circle complexity...

NewBie-image-Exp0.1 is the first experimental anime text-born graph model open-sourced by the NewBieAI-Lab team, using the Next-DiT architecture with 3.5B parameters, optimized for the secondary style. The model is optimized for the secondary style by a dual text encoder (GEMMA3-4B...

XVERSE-Ent是元象科技推出的专注于泛娱乐领域的开源大模型，包含中英文双版本，支持社交互动、游戏叙事和文化创作等场景。模型通过角色一致性强化、长剧情理解等技术优化，能在虚拟角色人设稳定性、复杂故...

VibeVoice-ASR是微软开源的统一语音转文本（ASR）模型，专为处理长音频设计，可一次性处理长达60分钟的连续音频，确保语义连贯性和说话人追踪的一致性。支持自定义热词功能，用户可输入特定词汇或...

Supertonic is open source, high-performance text-to-speech (TTS) system focused on rapid speech generation on local devices. Using ONNX Runtime technology, it can run on devices such as cell phones, computers and even Raspberry Pi, supports 23 languages and speech clones, and requires no network...

Mistral Vibe is an open source command line coding assistant from Mistral AI, developed based on the Devstral model, which supports natural language interaction to complete code search, file manipulation, version control and other tasks. Can automatically scan the project structure and Git status through the @ symbol...

Egocentric-10K is a large-scale first-person view (egocentric) factory operations video dataset open-sourced by the build.ai team. The dataset contains 10,000 hours of video, totaling 1.08 billion frames, involving 2...

NocoBase is based on AI-driven open-source no-code development platform that supports the rapid construction of business systems, without programming to complete the application development through configuration. The project uses Apache-2.0 protocol , provides private deployment and flexible scalability , suitable for enterprise management , collaboration platforms and other fields ...

Koina is an open source decentralized machine learning platform focused on simplifying proteomics data analysis. Developed by a team from the Technical University of Munich, Germany and the University of Michigan, USA. The platform integrates more than 30 mainstream models (e.g., ProSIT, MS²PIP) through a standardized interface and supports peptide mass...

Ming-flash-omni 2.0是蚂蚁集团开源的全模态大模型，集成了视觉、语音和生成能力的AI模型，在多项基准测试中性能领先。支持同时生成语音、环境音效和音乐，能通过自然语言指令精细控制音色、语...

MiniCPM-o 4.5 是面壁智能开源的 9B 全模态旗舰模型，以“边看边听主动说”的端到端架构，在手机端即可跑出 GPT-4o 级体验：支持单图、多图、高帧率长视频、实时语音双工对话，首 tok...

GLM-TTS is an open source industrial-grade speech synthesis system with powerful speech synthesis capabilities. Adopting a two-stage generation architecture: the first stage will be converted to text into speech token sequences, and the second stage will be converted into high-quality audio token sequences. The system supports only 3 seconds of voice samples to complete the sound...

FG-CLIP 2 is the world's leading graphical cross-modal visual language model (VL-M) launched by 360 Artificial Intelligence Research Institute, which surpasses similar models from Google and Meta in 29 authoritative benchmark tests, making it the most powerful VL-M at present.It is able to accurately recognize the gross...

Yume 1.5 is an open source interactive world generation model, jointly developed by Shanghai Artificial Intelligence Laboratory, Fudan University, and Shanghai Innovation Research Institute, which is capable of real-time interactive rendering (12 FPS on a single card). It adopts the joint spatio-temporal channel modeling (TSCM) technology, even if the context length increases...

Clawra是一个基于OpenClaw框架开发的AI女友程序，由韩国开发者David Im制作，具有完整人设和交互功能。通过Persona Engineering技术赋予AI“18岁亚裔女性练习生”的...

Devstral 2 is a family of next-generation programming models designed for software engineering tasks from Mistral AI, consisting of Devstral 2 (123B parameter) and Devstral Small 2 (24B parameter) versions.D...

InfinityStar is a unified spatio-temporal autoregressive framework open-sourced by ByteDance, designed for high-resolution image and video generation. Using a discrete autoregressive approach, it can simultaneously handle text-to-image, text-to-video, and image-to-video tasks in a single model. The framework is benchmarked in VBench ...

Vibe Kanban是开源的AI编程代理任务管理工具，专为同时使用多个AI编程助手（如Claude Code、Gemini CLI、Codex等）的开发者设计。通过看板形式统一管理任务进度，支持并行...

PersonaPlex是英伟达开源的全双工语音对话模型，拥有70亿参数。摒弃了传统的语音识别→语言模型→文本到语音的级联流程，采用统一的Transformer架构，能同步处理语音理解与生成。模型支持全...

Open-o3 Video is an open source video inference model jointly developed by Peking University and ByteDance, focusing on enhancing video inference through temporal and spatial evidence. By explicitly labeling key evidence with timestamps and bounding boxes, it helps the model better understand and interpret video content.

Petri is an open source AI security auditing framework developed by Anthropic that systematically assesses the security and behavioral alignment of AI models. By simulating a real-world scenario where an automated auditor engages in multiple rounds of conversations with a target model, followed by a judge agent that acts on the model's...

DeepOCR is an open source replication project that implements the core architecture of DeepSeek-OCR, which efficiently processes textual information through optical compression techniques. The core is DeepEncoder, consisting of SAM-base (processing high-resolution images), 16× convolutional compressor...

Xiaomi-Robotics-0 是小米开源的首代具身智能大模型，拥有47亿参数，采用"大脑+小脑"混合架构设计。视觉语言大脑基于多模态大模型，负责理解人类模糊指令与空间推理；动作执行小脑则通过Di...

Qwen3-VL-Reranker是阿里巴巴推出的多模态重排序模型，专门用于提升跨模态检索的精准度。与Qwen3-VL-Embedding协同工作：前者负责快速召回候选结果，后者通过深度跨模态交互（如...

ROCK (Reinforcement Open Construction Kit) is Alibaba's open source sandbox for training environment of intelligences, which solves the problem that intelligences can't be scaled up for training in real environments.ROCK provides a highly stable sandbox management service...

Seekdb (OceanBase Seekdb) is Ant OceanBase open source AI native hybrid search database, supporting unified hybrid search of vector, full text, scalar and geospatial data, using multi-stage retrieval mechanism, to achieve high precision search under low latency ...

SAIL-VL2 is an open source multimodal visual language model by the Byte Jump team, focusing on joint modeling of multimodal inputs such as images and text. Using the sparse mixture of experts (MoE) architecture and progressive training strategy, it achieves high performance at parameter scales from 2B to 8B, especially in the areas of graphic comprehension, math...

GigaBrain-0 is the first end-to-end Vision-Language-Action (VLA) embodied base model in China that uses world model generation data to realize real machine generalization, and it is jointly released as open source by GigaVision and Hubei Humanoid Robot Innovation Center. It adopts the hybrid Transformer architecture, integrating ...

UltraEval-Audio是清华大学NLP实验室、OpenBMB和面壁智能联合开发的音频模型评测框架，最新版本为v1.1.0。专注于解决音频模型复现难、依赖冲突等问题，提供一键复现热门模型（如Vo...

FLUX.2 is an open source image generation and editing model released by Black Forest Labs that supports textual raw images, multi-image referencing, and image editing with richer details, clear textures, and stable lighting. There are four versions: FLUX.2 [pro] (comparable to the top closed source...

SoulX-FlashTalk是Soul App AI团队开源的实时数字人生成模型，拥有140亿参数量，实现了0.87秒超低延迟和32帧/秒的高帧率。模型通过双向蒸馏技术解决了传统数字人延迟高、画面易...

PromptFill is a structured cue generation tool designed specifically for AI painting, which helps users quickly build, manage and iterate complex prompts through visual "fill-in-the-blank" interactions, improving the efficiency and quality of AI image generation.PromptFill's core features...

Bee is a full-stack open-source multimodal big model solution jointly launched by Tencent Mixed Element team and Tsinghua University to narrow the performance gap between open-source and closed-source models by improving data quality. The project contains three core achievements: 15 million scale high-quality two-layer CoT dataset Honey-Data...

Kimi K2.5 是月之暗面发布的开源旗舰模型，采用 1T MoE 架构、激活 32B、上下文 256K token，原生支持图文视频多模态输入。在 Agent、代码、视觉理解三大基准均列开源第一...

KoalaQA is an open source intelligent after-sales service system developed by the Chaitin team. Based on the AI model, it provides AI customer service, AI search and knowledge base management functions to help enterprises quickly build an intelligent Q&A platform. The system supports 24/7 real-time response ...

Protenix-v1是字节跳动ByteDance Seed团队推出的首个开源蛋白质结构预测模型，性能在严格对齐训练数据和模型规模后超越AlphaFold 3。模型具备显著的推理时扩展特性：通过增加采...

GLM-4.6V is a series of multimodal large language models open-sourced by Smart Spectrum AI. The series contains two versions: GLM-4.6V (106B-A12B), the basic version for cloud and high-performance cluster scenarios, with the Mixed Expert (MoE) architecture, a total of about 106 billion references, and an activation...

openPangu-VL-7B是华为开源的7B参数规模的多模态模型，专为昇腾端侧设备优化设计。模型在视觉定位、OCR识别、文档理解等任务中表现出色，支持实时推理（5FPS），单卡延迟仅160毫秒。

opcode is designed for Claude Code open source graphical desktop application , the developer winfunc based on Tauri 2 + React 18 + Rust development. Provides a visual interface to manage Claude Code projects , support for creating ...

PartCrafter is an advanced 3D generative model, jointly proposed by Peking University, ByteDance and Carnegie Mellon University. It can generate multiple semantically explicit and geometrically diverse 3D mesh parts from a single RGB image at once. The models are modeled through a combinatorial potential space and...

Ring-2.5-1T 是蚂蚁集团百灵大模型团队开源的全球首个万亿参数混合线性架构思考模型，采用1:7 MLA与Lightning Linear Attention混合设计，激活参数量达63B。模型在...

Infographic is a new generation of Ali AntV team open source framework , based on G2 and Ant Design development , focusing on rapid generation of high-quality infographics , providing 30 + layout templates , 120 + preset themes and AI intelligent generation capabilities .

Vidi2 is a second-generation multimodal video understanding and generation big model open-sourced by ByteDance, focusing on video content understanding, analysis and creation. It supports joint input of text, video, and audio modalities, and can simultaneously understand picture content, sound information, and natural language commands to achieve cross-modal interaction and push...

Fun-Audio-Chat-8B is an open source 8 billion parameter end-to-end speech big model by Ali Tongyi team, direct speech in speech out, no need for ASR+LLM+TTS splicing, bilingual fluent in Chinese and English, with low latency and natural timbre. Using dual-resolution shared LLM with 25Hz...

Ming-Omni-tts 是蚂蚁集团与Inclusion AI联合开源的多模态音频生成模型，包含0.5B和16.8B-A3B两个版本。模型首次实现了语音、环境音和音乐的统一自回归生成，支持语速、音量...

Computer Use Preview is Google's open source AI browser automation tool based on the Gemini model , through natural language commands to achieve web page interaction . Using "screenshot→analysis→execution" visual recognition process , support Playwrigh...

Wujie-Emu3.5 is an open source multimodal world grand model from Beijing Zhiyuan Artificial Intelligence Research Institute, with 34 billion references and native world modeling capability. Trained by 10 trillion multimodal Token (including 790 years of video data), it can simulate the laws of physics and realize graphic generation, visual guidance...

RealVideo is an open-source real-time streaming video generation system from Smart Spectrum AI that can quickly generate natural and smooth video responses in 2 to 3 seconds. Users only need to upload a photo and enter text, and the system can generate corresponding voice and video, realizing real-time conversations with AI characters...

Chroma 1.0是FlashLabs发布的全球首个开源的实时端到端语音对话模型，兼具低延迟交互、高保真个性化语音克隆和强对话能力。通过紧密耦合语音理解与生成，采用1:2文本-音频token调度策略...

VibeVoice-Realtime is Microsoft's open source lightweight real-time text-to-speech (TTS) model designed for low latency and real-time interaction. Supports streaming text input , from the first text token can be vocalized , the delay is only about 300 milliseconds , suitable for dynamic number ...

SkyReels-V3是昆仑万维Skywork AI开源的多模态视频生成模型，被誉为视频生成领域的"全能型"标杆。模型基于"一核多支"的统一架构，在单一建模框架内集成三大核心能力：参考图像转视频、智能...

BestBlogs is a platform focused on providing high-quality content for technology practitioners, entrepreneurs, product managers, and more. Through RSS feeds and crawler technology, it collects articles, podcasts, videos, and other multi-format content from more than 400 high-quality blogs. The core strength lies in utilizing AI big language...

InkSight is Google's open source AI handwriting recognition tool that converts paper handwritten notes into editable digital inked files (e.g. SVG format). Unlike traditional OCR , can recognize text content , can restore the handwriting style , paragraph structure and focus marking , support for multi-language processing .

ClawWork是香港大学数据科学实验室开发的AI经济压力测试框架，允许AI在模拟经济环境中完成真实工作任务并获得报酬。核心逻辑是让初始资金仅10美元的AI通过完成220个专业任务（覆盖制造、金融、医...

New API是基于Go语言开发的开源AI聚合网关工具，可统一管理30+种主流大模型（如OpenAI、Claude、Midjourney等），将不同模型接口转换为标准化OpenAI格式。

Step3-VL-10B是阶跃星辰团队开源的100亿参数多模态AI模型，核心突破在于以轻量化设计实现顶级性能。模型通过统一预训练策略（1.2T多模态令牌数据）和创新的并行协同推理技术（PACORE...

LingBot-World 是蚂蚁集团旗下具身智能公司灵波科技（Robbyant）开源的交互式世界模型，专为具身智能、自动驾驶及游戏开发打造高保真“数字演练场”。模型通过可扩展数据引擎从大规模游戏环境...

GigaWorld-0 is the open source world modeling framework of domestic Embodied Intelligence startup GigaAI, mainly used to solve the data bottleneck problem in the field of Embodied Intelligence (Embodied AI). Efficiently generating high-quality, diverse and physically realistic training data, push...

PaperBanana是北大与谷歌团队联合开源的AI学术插图自动生成框架，专门解决科研人员绘制方法示意图和统计图表的痛点。框架通过五个智能体协作（检索、规划、造型、渲染和批评），实现从文本描述到Neu...

LingBot-Depth是蚂蚁灵波科技开源的高精度空间感知模型，专门解决机器人在透明玻璃、反光物体等复杂场景中的深度识别难题。模型通过创新的"掩码深度建模"技术，在RGB图像基础上预测缺失的深度值

StoryMem是字节跳动与南洋理工大学联合开源的AI视频生成系统，专为解决多场景视频中角色和环境一致性问题。核心通过"视觉记忆库"技术，自动存储关键帧并在后续生成时参考，确保人物外貌、服装、场景元素...

Youtu-LLM 是腾讯 Youtu 团队开源的轻量级语言模型，参数规模为 19.6 亿。专为智能体任务设计，具备强大的“原生智能体能力”，在多项任务中超越同规模甚至更大模型。

Flowra is ModelScope joint woo mile WULI team open source graph execution engine and node package development tools, is the core component of FlowBench. Through the directed acyclic graph (DAG) organization workflow , with intelligent caching , parallel scheduling , distributed support ...

SenseNova-SI is an open source spatial intelligence grand model released by ShangTech, focusing on improving AI's ability in spatial understanding and reasoning. The model excels in six core dimensions, including spatial measurement, reconstruction, relationship judgment, perspective transformation, deformation analysis, and spatial reasoning, significantly outperforming other...

QwenLong-L1.5 is an open source long text inference model from Alibaba Tongyi Lab, focusing on solving complex inference problems with ultra-long contexts (e.g., 1M-4M tokens). The core breakthrough lies in three major innovations in the post-training phase: through knowledge graph, SQL parsing and multi-intelligence...

VoiceSculptor 是西北工业大学联合多家机构开源的音色设计模型，基于 LLaSA-3B 和 CosyVoice2 开发，专注于通过自然语言指令生成多样化音色的语音合成。支持对语速、音量、基频...

AgentCPM-Report 是清华大学自然语言处理实验室、中国人民大学、面壁智能与 OpenBMB 开源社区联合研发的深度调研智能体工具。基于 8 亿参数的模型，通过深度检索和推理，能生成万字长篇...

LingBot-VA 是蚂蚁灵波开源的全球首个“自回归视频-动作世界模型”，把视频生成与机器人控制塞进同一 Transformer，每一步同时输出下一帧世界画面和对应动作，实现“边想边干”。

GLM-4.7-Flash是智谱开源的混合专家架构语言模型，参数规模为30B，激活参数量3B，上下文窗口达200K，最大输出令牌为128K。在编程能力上表现出色，SWE-bench验证集分数达59.2...

RoboCOIN is the world's first large-scale dual-arm robot real machine dataset open-sourced by Beijing Zhiyuan Artificial Intelligence Research Institute in conjunction with a number of enterprises and colleges and universities, which contains 15 types of robot platforms, 180,000 real operation trajectories, and 421 types of task scenarios. The most important feature is the use of hierarchical annotation system to disassemble the task ...

LongCat-Image is an open source image generation and editing model released by the LongCat team of Meituan. Using a hybrid backbone architecture (MM-DiT+Single-DiT), combined with a visual language model (VLM) conditional encoder, it is able to realize text-generated images and multiple rounds of image editing...

Mistral 3 is the latest multimodal large model series released as open source by Mistral AI, including the flagship model Mistral Large 3 (675B total parameters) and the lighter version of the Ministral series (3B/8B/14B), both supporting image understanding...

Nemotron Speech ASR是英伟达开源的实时语音识别模型，专为低延迟场景优化，支持24毫秒极速转录和多人并发对话。核心采用混合Mamba-Transformer MoE架构，通过固定状态缓...

ChatTutor is an open source AI teaching aid focused on visual and interactive learning of STEM subjects. Through the multi-intelligent body architecture to achieve dialogical Q&A and dynamic drawing function, can draw math graphs, physics circuits or mind maps on the whiteboard in real time, to help users intuitively understand the abstract generalization ...

UnifoLM-VLA-0 是宇树科技 UnifoLM 系列的首款操作型大模型，突破传统视觉语言模型（VLM）仅能理解图像文字的局限，通过在机器人操作数据上的持续预训练，实现从"图文理解"向具备物理常...

Ovis-Image is a 7 billion parameter text-generated graph model open-sourced by the AIDC-AI team of Alibaba International Digital Commerce Group, focusing on high-quality text rendering. Based on Ovis-U1 architecture, it inherits the advanced visual decoder and bi-directional Token refiner ...

NovaSR是开源的音频超分辨率模型，主要用于将低质量音频（如16kHz采样率的电话音质）提升为高质量音频（如48kHz采样率的录音室级音质）。模型大小仅52KB，比一张微信表情包还小，可轻松部署在资...

Genie Sim 3.0是智元机器人发布的首个大语言模型驱动的开源仿真平台。基于NVIDIA Isaac Sim构建，融合三维重建、视觉生成技术与物理引擎，实现毫米级精准复刻真实环境，通过自然语言指...

GLM-Image是智谱与华为联合开源的多模态图像生成模型，基于昇腾Atlas 800T A2芯片和昇思MindSpore框架训练，采用创新的"自回归+扩散解码器"混合架构。核心突破在于实现了国产芯片...

Step 3.5 Flash 是阶跃星辰开源的 1960 亿稀疏 MoE 模型，每 token 仅激活 110 亿参数，能在代码任务跑出 350 token/s 的实时速度。基于自研 MTP-3 多 ...

MOVA（MOSS-Video-and-Audio） 是上海创智学院 OpenMOSS 团队联合模思智能（MOSI）开源的端到端音视频生成模型，是中国首个高性能开源音视频模型。突破了传统"先画面后配音...

Baichuan-M3是百川智能推出的新一代开源医疗大语言模型，专为医疗场景深度优化，具备强大的医疗推理和问诊能力。在权威的HealthBench评测中以65.1分的综合成绩位列全球第一，超越了GPT...

MiniMax M2.1 is MiniMax's open source coding and agent model with 10 billion activations and support for many major programming languages such as Rust, Java, Golang, C++, Kotlin, Objective-C, TypeS...

Intern-S1-Pro是上海人工智能实验室开源的全球首个万亿参数级科学多模态大模型。采用512专家MoE架构，激活仅8专家22B参数，兼顾性能与效率。模型基于SAGE架构，引入傅里叶位置编码，统一...

SenseNova-MARS 是商汤开源的首个支持动态视觉推理与图文搜索深度融合的智能体视觉语言模型（Agentic VLM），提供 8B 和 32B 双版本。模型能自主规划任务步骤、调用多种工具（如...

OctoCodingBench是MiniMax开源的首个面向Coding Agent生产级标准的评测集，核心创新在于通过Check-level准确率(CSR)和Instance-level成功率(IS...

TranslateGemma是谷歌推出的基于Gemma 3的开源机器翻译模型系列，专为提升翻译质量而设计。通过两阶段微调（监督微调和强化学习）优化翻译效果，提供4B、12B、27B三种参数规模，支持5...

RynnBrain 是阿里巴巴达摩院开源的具身智能大脑基础模型，为机器人提供深度环境理解和物理世界交互能力。是业界首个赋予机器人时空记忆和物理空间推理能力的开源模型。包含2B、8B、30B等7个不同参...