Xiaomi MiMo Token Plan Global Launch
Xiaomi has launched the Xiaomi MiMo Token Plan subscription, adopting a unified Credit billing system and removing the 5-hour token limit, with four tiers available to support mainstream AI development tools.
88Z AI / NEWS
The news, the context, and what changes.
Xiaomi has launched the Xiaomi MiMo Token Plan subscription, adopting a unified Credit billing system and removing the 5-hour token limit, with four tiers available to support mainstream AI development tools.
AI video platform OiiOii has officially opened registration, integrating the full-power Seedance 2.0 model for high-quality video generation without queues. The update introduces multi-panel storyboard previews, a free canvas mode, and scene asset features, along with new styles like realistic human portraits, significantly boosting creation efficiency and visual consistency.
Alibaba Tongyi Lab releases the Wan2.7-Video model, supporting full-modal input and instruction-based video editing, with capabilities for subject consistency control, storyboard segmentation, and intelligent plot design.
Zhipu has released GLM-5V-Turbo, a multimodal coding foundation model that natively integrates visual and text capabilities. It supports generating runnable code directly from design drafts or screenshots, with a context window of 200k. The model leads in multimodal coding and GUI Agent benchmarks and is deeply adapted to frameworks like Claude Code and AutoClaw, enabling AI agents with visual perception.
Tongyi Lab releases Qwen3.6-Plus, accessible via Alibaba Cloud Bailian API. The model focuses on coding agent capabilities, supports million-level context, enhances multimodal reasoning, and leads in code generation, repair, and front-end development tasks.
OpenClaw has officially launched the ClawHub mirror site, with infrastructure support from ByteDance's BytePlus and Volcano Engine, aiming to resolve slow access to the skills market in China. The site offers a complete Chinese interface and data synchronization, continuously curating high-quality community Skills. Users can access it directly via the specified domain and seamlessly invoke various Agent skills within OpenClaw.
Bilibili's self-developed AI tool updream has entered beta testing, offering invited creators lightweight intelligent services including inspiration generation, smart editing, and project management.
OpenAI closed a record $122 billion financing round, reaching an $852 billion valuation with lead investors including Amazon and Nvidia. Due to high costs and low retention, Sora was shut down as the company pivots to text, code, and enterprise services to prepare for an IPO.
Tongyi Lab has released the Wan2.7-Image model, focusing on human realism, text stability, and color accuracy. The model supports deep customization of virtual avatars to avoid homogeneous faces, stably generates images with ultra-long text including multiple languages, tables, and formulas, and introduces a Hex value palette for precise control of brand colors.
The Meituan LongCat team has released the LongCat-AudioDiT model, achieving state-of-the-art performance in zero-shot voice cloning. The model performs diffusion generation directly in the waveform latent space, discarding traditional Mel-spectrogram intermediate representations to prevent information loss. It also introduces Dual Constraint Alignment and Adaptive Projection Guidance techniques to address training-inference mismatches and mitigate oversaturation.
AIsphere announced three major updates for PixVerse on the second day of its Lightning Release Week: a Quick App, a Team Edition, and a CLI tool. The Quick App debuts with 'Ad Master,' enabling rapid conversion of product images into ad videos. The Team Edition supports collaboration for 2 to 100 users, offering shared credits, four-level permission management, and independent spaces. The CLI tool allows developers to invoke video generation capabilities via command line, facilitating integratio
Kunlun Wanwei has launched the Tiangong Short Drama Workbench, integrating four major video models and multi-agent technology to provide a one-stop industrial creation process from script to final film.