The high-speed version of Kimi K2, dubbed 'kimi-k2-turbo-preview,' has launched with a 4x performance boost, now delivering 40 tokens per second. A limited-time 50% discount is available until September 1, offering competitive pricing for input and output tokens. The team plans further optimizations to enhance speed and user experience.
Tsinghua University's MOSS-TTSD, a cutting-edge text-to-speech model, has been open-sourced. Trained on over 1 million hours of voice data, it excels in dialogue generation for podcasts and audiobooks. The model features innovative voice cloning and supports ultra-long speech generation up to 960 seconds.
Deep Cogito's Cogito v2 series introduces 'machine intuition' and self-improving reasoning, challenging established models like DeepSeek and Claude. The open-source models, ranging from 70B to 671B parameters, offer shorter reasoning paths and lower training costs. The flagship Cogito v2-671B MoE demonstrates superior efficiency in mathematical and legal reasoning tasks.
ByteDance's Seed team has launched Seed Diffusion Preview, an experimental diffusion language model that achieves 5.4x faster inference speeds than comparable autoregressive models while maintaining competitive performance on code generation tasks. The model introduces four key innovations to overcome challenges in applying diffusion techniques to discrete language data.
Black Forest Labs and Krea have open-sourced FLUX.1-Krea, an AI image generation model designed to eliminate artificial traces in AI-generated visuals. The model emphasizes natural aesthetics through advanced training techniques, including human feedback reinforcement learning, and is fully compatible with the existing FLUX ecosystem.
Augment has unveiled Auggie, a new CLI tool designed to streamline developer workflows with AI-driven automation. The tool integrates seamlessly with popular IDEs and offers advanced context-aware features, positioning itself as a strong competitor in the enterprise AI coding space.
Wanxi Technology's Tianmu 2.0 model secured fourth place in SuperCLUE's July Text-to-Video Large Models ranking, showcasing its AI video generation prowess. The company also announced a collaboration with Huawei Cloud to establish an AI Video Lab, aiming to innovate digital creative solutions.
Manus, an AI agent platform, is set to release its Wide Research feature, enabling parallel processing with multiple AI agents. This update aims to handle complex tasks like large-scale data analysis and design generation. The feature will debut for premium users at $199/month, marking a significant advancement in AI-driven research tools.
AI healthcare startup C8Health has raised $12 million in Series A funding to address medical information silos. Its platform integrates clinical guidelines across hospital systems, providing real-time access to critical knowledge. The solution, already deployed in 100+ US hospitals, achieves 90% clinician adoption within months by delivering context-aware recommendations.
Alibaba Cloud's Tongyi Lab has open-sourced its AI agent project WebAgent, featuring flagship components WebShaper and WebSailor. The system demonstrates human-level web interaction capabilities, outperforming competitors in GAIA evaluations while offering applications in academic research and business analysis.
Midjourney has launched a 'Recommended for You' feature, offering users AI-generated images and videos tailored to their preferences. The one-click tool analyzes user interactions to deliver customized content, enhancing productivity and personalization. The feature supports style and chaos parameters for further refinement.
Meta has rolled out a significant update to its Avatar system, introducing multiple preset body types, detailed customization options, and AI-powered styling. The update includes new poses, expressions, and fashion features, aiming to provide users with more personalized virtual representations. AI styling is currently available in the U.S. and Canada, with plans for global expansion.
Ollama has released a desktop client, replacing its command-line interface with a user-friendly GUI. The new version supports drag-and-drop document processing, multimodal recognition, and local AI model management. Available initially for macOS, it enhances accessibility while maintaining privacy-focused local operations.
The OWL team has open-sourced Eigent, a revolutionary multi-agent collaboration tool designed to enhance efficiency in complex task processing. Built on the OWL framework, Eigent leverages parallel processing and dynamic customization to streamline workflows. With features like Human-in-the-Loop and extensive tool integration, it promises significant advancements in AI task automation.
The OWL team has open-sourced Eigent, a multi-agent collaboration tool built on the OWL framework. Designed for complex task automation, Eigent enables parallel processing of sub-tasks by specialized agents, improving efficiency and scalability. The tool integrates with major AI models and supports local or cloud deployment, marking a significant advancement in open-source AI ecosystems.
Alibaba Cloud has launched 'No Shadow AgentBay,' a cloud platform designed to streamline AI agent development and operation. The platform offers high-performance computing, multi-OS support, and advanced AI capabilities, enabling developers to handle complex tasks with ease. Security and scalability are key features, making it a robust solution for modern AI applications.
The Qwen3-30B-A3B model has released a new version, Qwen3-30B-A3B-Instruct-2507, achieving significant improvements in general capabilities, multilingual support, and long-text understanding. The update positions it competitively against top closed-source models like Gemini2.5-Flash and GPT-4o.
The Shenzhen Institute of Advanced Technology has unveiled the HYPIR Image Restoration Large Model, capable of transforming old photos into 8K ultra-high-definition images in just 1.7 seconds. This breakthrough promises significant advancements in cultural heritage preservation, film restoration, and medical research.
Google has launched a new Video Overviews feature for its AI note-taking app, NotebookLM, designed to simplify complex information through slides and voiceovers. The tool extracts data from user-uploaded materials to create engaging visual summaries. Initially available in English, Google plans to expand support for multiple languages and refine content presentation.
Google has enhanced its Imagen 4 model, now ranking among the top three in AI image generation. The upgraded Imagen4Ultra offers improved quality, faster speeds, and competitive pricing, positioning it as a strong rival to OpenAI's GPT-4o and Seedream3.0.
