Flash

The news, the context, and what changes.

RSS
Flash

Xiaohongshu Releases First 'AI Governance Principles'

Xiaohongshu has released its first AI governance principles, establishing three core tenets: supporting AI as a creative aid, requiring labeling of AI-generated content either by users or the platform, and strictly prohibiting AI-driven fraud and infringement. The platform will enhance its detection capabilities and open an appeal channel to protect creator rights and maintain community authenticity and fairness.

88Z AI editorial1 min read
Flash

OpenAI and Microsoft Announce Formal Removal of Exclusive Cloud Partnership Restrictions

Microsoft and OpenAI have revised their agreement, removing exclusive cloud partnership restrictions. While OpenAI products remain prioritized on Azure, services can now be offered to customers via other cloud providers. Intellectual property licensing has shifted from exclusive to non-exclusive, extended until 2032. Microsoft will no longer pay revenue shares to OpenAI, while OpenAI's payments to Microsoft are capped and decoupled from technical progress. Microsoft also gains the right to indep

88Z AI editorial1 min read
Flash

Baidu Wenku and Netdisk Jointly Launch General Agent GenFlow 4.0

Baidu launched the general agent GenFlow 4.0 at AI DAY, with monthly active users exceeding 100 million. The new version upgrades the Office Agent to support one-click PPT generation, Excel processing, and Word document writing, while enhancing memory center capabilities. It also deeply integrates with OpenClaw for seamless cross-device collaboration and skill migration.

88Z AI editorial1 min read
Flash

Alibaba Tongyi Launches New Reinforcement Learning Framework EAPO

Alibaba Tongyi Lab has released EAPO, a reinforcement learning framework that optimizes long-text reasoning through an evidence reward mechanism to address model hallucinations. Tests based on Qwen3-30B show its performance surpassing larger mainstream models like GPT-OSS and Claude-Sonnet-4.

88Z AI editorial1 min read
Flash

Alibaba QoderWork Launches Expert Suite Feature

Alibaba QoderWork launches the Expert Suite, featuring ten pre-built kits across six sectors including finance and law, supporting one-click usage and custom encapsulation.

88Z AI editorial1 min read
Flash

LibTV Premieres HappyHorse 1.0 Video Model

LibTV has premiered the new HappyHorse 1.0 video model, supporting text-to-video, image-to-video, and reference image generation. It produces 15-second 1080P multi-shot narrative videos with features like natural language editing, cinematic quality, smart storyboarding, audio-visual sync, and style restoration, significantly improving character motion, micro-expressions, and dialogue realism.

88Z AI editorial1 min read
Flash

Duiyou Launches HappyHorse 1.0 Video Model

Alibaba's Duiyou platform launches the HappyHorse 1.0 video model, supporting scenarios like 3D animation and drone shots, with limited-time discounts as low as 12 Duidou per second.

88Z AI editorial1 min read
Flash

DeepSeek API Input Cache Price Cut by 90%

DeepSeek officially announced that the price for input cache hits across all API services has been reduced to one-tenth of the original rate, aiming to lower developer costs and improve cost-effectiveness in long-context scenarios.

88Z AI editorial1 min read
Flash

Xiaomi Open-Sources Full Post-Training Pipeline for VLA Model Xiaomi-Robotics-0

Xiaomi's Robotics Division has open-sourced the full real-device post-training pipeline for Xiaomi-Robotics-0. Built on a pre-trained base, the model achieves high-precision tasks like earphone storage with just 20 hours of task data. The team employed strategies such as asynchronous inference, action prefixes with adaptive weighting, Lambda-type masking, and random masking to address the model's 'laziness effect,' balancing action coherence and visual sensitivity.

88Z AI editorial1 min read
Flash

Lingguang App Launches World Model Experience Feature

Lingguang App has launched the 'Experience World Model' feature, bringing world model capabilities to mobile devices for the first time. Users can generate interactive 3D virtual worlds from images and explore them immersively in first-person view.

88Z AI editorial1 min read
Flash

Xingdong Jiyuan Completes New Funding Round Exceeding $200 Million

Embodied AI company Xingdong Jiyuan has completed a new funding round exceeding $200 million, led by SF Group with participation from Sequoia China, IDG Capital, and CICC. The company has achieved the industry's first product-market fit, deployed solutions in multiple logistics centers with efficiency surpassing humans by 85% in some scenarios, and plans to begin mass delivery of thousands of units in Q2 2026.

88Z AI editorial1 min read
Flash

StepFun and Tencent Cloud Establish Strategic Partnership

StepFun and Tencent Cloud have formed a strategic partnership to integrate voice large models and multimodal technologies with Tencent Cloud's infrastructure advantages, accelerating the adoption of large models in vehicles. The collaboration focuses on smart cockpits, creating proactive AI interactions featuring wake-word-free operation, continuous dialogue, and emotion recognition, while linking with Tencent Music, video, and map services to achieve a closed-loop in-car service ecosystem.

88Z AI editorial1 min read