Flash

The news, the context, and what changes.

RSS
Flash

LiblibAI Launches Overseas AI Video Creation Platform Newtake

LiblibAI has launched Newtake, an overseas AI video creation platform focused on cinematic content. Leveraging an infinite canvas and node-based workflow, it integrates professional features like a director's console and character consistency management, while supporting the Seedance 2.5 model to lower barriers for high-end video production.

88Z AI editorial1 min read
Flash

China Telecom Tianyi AI Launches Desktop General Agent Assistant TeleAgent

China Telecom Tianyi AI has launched the desktop version of its Xingchen Super Agent, TeleAgent, positioned as an AI office companion. It supports Windows 10 and MacOS 13 or later. The tool automates tasks such as file organization, Excel summarization, data analysis, PPT creation, and information gathering, allowing users to execute commands via voice or text.

88Z AI editorial1 min read
Flash

Alibaba Officially Launches Video Generation Model Wan3.0

Alibaba Cloud announces the official launch of video generation model Wan3.0, supporting single-generation of 30-second videos and input from various document formats, with applications deployed across multiple industries.

88Z AI editorial1 min read
Flash

RunningHub Launches Next-Generation Video Generation Model Wan3.0

The Wan3.0 model has officially launched on the RunningHub platform. This model supports single-generation videos up to 30 seconds long and accepts inputs including text, images, audio, video, and various document formats, enabling one-click conversion of PPTs into videos. It demonstrates significant improvements in character and scene consistency, effectively addressing the 'AI face' issue. Currently, RunningHub's ecosystem products, such as rhTV, RHSTORY, and ComfyUI, have been fully adapted t

88Z AI editorial1 min read
Flash

Alibaba Tongyi Launches Qwen-UI-Agent, a Foundation Model for GUI Agents

Alibaba Tongyi has released Qwen-UI-Agent, a foundation model for GUI agents designed for mobile, desktop, and web scenarios. The model outperforms leading flagship models in various benchmarks and features strict safety confirmation mechanisms to ensure secure user operations.

88Z AI editorial1 min read
Flash

OpenAI Open-Sources Codex Harness, the Underlying Execution Framework for AI Agents

OpenAI has officially open-sourced Codex Harness, its underlying execution framework for AI agents. The framework covers the complete execution loop, including task understanding, memory retention, tool invocation, and human approval. It supports integration into business systems via CLI, SDK, and app-server. Key features include sandboxed secure execution, native Human-in-the-Loop mechanisms, streaming event transmission, and a model-agnostic architecture, aiming to provide developers with a fl

88Z AI editorial1 min read
Flash

SenseTime Open-Sources Native Unified Multimodal Large Model SenseNova U1.5 Lite

SenseTime has officially open-sourced the lightweight native unified multimodal large model, SenseNova U1.5 Lite. The model focuses on real-world visual creation workflows, with enhanced capabilities in long instruction following, visual quality, text and layout control, native 4K high-resolution output, and precise visual editing.

88Z AI editorial1 min read
Flash

Ling-3.0 tiny & flash Base Models Officially Open-Sourced

Ant Group's Bailin large model open-sources Ling-3.0-tiny-base and Ling-3.0-flash-base, releasing weights for six key training stages. The Tiny version focuses on code and low-cost research, while the Flash version balances code, reasoning, and long-context capabilities.

88Z AI editorial1 min read
Flash

Luming Robotics Launches Embodied AI Evolution Engine Lumos NexCore

Luming Robotics unveiled the Lumos NexCore embodied AI evolution engine at the 2026 World Robot Conference. Positioned as industrial 'skill infrastructure,' it integrates data, training, evaluation, encapsulation, and operations to enable cloud-like invocation and iteration of robot capabilities.

88Z AI editorial1 min read
Flash

DeepSeek Harness Update Enhances Multimodal Capabilities

DeepSeek Harness releases version v0.1.0-rc.8, focusing on enhanced multimodal support with native image requests and mixed text-image inputs, while allowing Claude Code and Codex to be called as sub-agents. The update also improves tool concurrency and Windows terminal experience, fixes stability issues, and reinforces its role as a unified agent scheduling layer.

88Z AI editorial1 min read