Flash

The news, the context, and what changes.

RSS
Flash

Alipay and OPPO Partner for Cross-Device Agent Interconnectivity

Alipay and OPPO have partnered to enable cross-device agent interconnectivity, integrating Alipay's AI agent 'Abao' into OPPO's assistant 'Xiaobu'. Users can now use voice commands to let Xiaobu invoke Abao for nearly 200 life services, such as topping up phone credit and buying tickets, without opening apps.

88Z AI editorial1 min read
Flash

Amap Launches General World Model Workshop ABot-World Studio

Amap has launched ABot-World Studio, a general world model workshop that unifies interactive video generation with 3DGS scene creation for the first time. Users can create AI worlds that are real-time interactive and freely shareable using text or image inputs. The tool supports local deployment on a single 5090 GPU, maintains stable continuous inference for over an hour, and features a space-time door for cross-scene transitions.

88Z AI editorial1 min read
Flash

AIsphere Completes 2.98 Billion Yuan Series C Financing

AIsphere announced the completion of its overall Series C financing, totaling 2.98 billion yuan. The Series C+ round was led by Alibaba, with participation from over ten domestic and international institutions. Funds will be used for R&D of video generation foundation models and real-time world models, as well as global product growth. Its PixVerse platform has over 150 million global users. Earlier this year, it launched PixVerse R1, the world's first general real-time world model supporting 10

88Z AI editorial1 min read
Flash

StepFun Launches First LLM-Native Agent Phone STEPX Neo

StepFun has released the STEPX Neo, its first AI agent phone featuring the self-developed Step AOS and personal agent Amoo. It supports natural language cross-app scheduling and cloud-edge synergy, with initial integration of major apps like Alipay and Meituan.

88Z AI editorial1 min read
Flash

Looki Completes Hundreds of Millions in A1 Round Financing

Guangzhi Shikong (Looki) announced the completion of hundreds of millions in A1 round financing, led by Yuansheng Capital, with existing shareholders such as BAI Capital increasing their stakes. The company focuses on the personal intelligent wearable sector, achieving a technical upgrade from passive response to active prediction through its PIE Active AI Engine 2.0. By combining scene understanding with long-term memory, it anticipates user needs. Its first multimodal AI hardware, Looki L1, is

88Z AI editorial1 min read
Flash

LibTV Launches Professional Video Creation Agent LibTV Agent

LibTV, under Libu Libu AI, has launched LibTV Agent, the world's first professional video creation agent. It features four core capabilities: a high-baseline all-around agent, the largest professional video Skill Hub, dual-view creation, and agent-driven editing. Over 100 director-level Video Skills are available initially, covering film ads, short dramas, and self-media, with an over 80% success rate for videos under three minutes.

88Z AI editorial1 min read
Flash

StepFun Launches Step Edge On-Device Model Suite

StepFun has released Step Edge, an on-device model suite for terminals like phones and cars, covering base, audio, GUI, and generation models. With native cloud-edge synergy, it achieves 0.1s low local latency and full-modal privacy protection, ranking first in 29 core benchmarks across text-vision, audio, GUI, and image generation.

88Z AI editorial1 min read
Flash

SenseTime Open-Sources SenseNova-Vision Unified Visual Large Model

SenseTime has officially open-sourced the SenseNova-Vision unified visual large model, natively integrating classic tasks like detection and segmentation into the large model framework, demonstrating robust capabilities in zero-shot generalization and complex scene understanding.

88Z AI editorial1 min read
Flash

Tencent Hunyuan Open-Sources End-to-End OCR Large Model HyOCR-1.5

Tencent Hunyuan has released HyOCR-1.5, the industry's first fully open-source end-to-end OCR expert large model for both training and inference. By introducing DFlash speculative decoding, the model significantly accelerates inference speeds and ranks first in end-to-end tasks on the OmniDocBench v1.6 benchmark with a score of 94.74. HyOCR-1.5 supports eight types of tasks, including document parsing, ancient character recognition, and multi-page Q&A, covering 331 languages.

88Z AI editorial1 min read
Flash

OpenAI Codex, ChatGPT Work Temporarily Remove 5-Hour Usage Limit

The head of OpenAI Codex engineering announced that the five-hour usage limit for Plus, Business, and Pro subscribers has been temporarily lifted for Codex and ChatGPT Work. Additionally, the GPT-5.6 Sol model will improve efficiency, allowing users to complete more tasks within the same quota.

88Z AI editorial1 min read