SponsoredAI 会员在线充值
Visit sponsor Chatgpt/Claude/Grok 会员24h在线自助充值 安全快速 稳定保障
奥创(Ultra)商城会员充值中心Loading…
SponsoredChatgpt/Claude/Grok 会员24h在线自助充值 安全快速 稳定保障
奥创(Ultra)商城会员充值中心
Sponsored本页公开展示实时样本、模型覆盖与响应时间。先用 5 元按量跑自己的 Claude Code/Codex 任务，成功后再决定是否长期使用。
ACS Gateway
Sponsored【腾讯云】WorkBuddy 全场景 AI 办公工作台，团队旗舰版低至198元/月
腾讯云
Sponsored人人可用的AI应用创作平台 自然语言交互，全栈应用搭建，伟大创意快速实现;个人版新用户注册送5000积分
阿里云Source-reviewed updates on AI models, APIs, research, security, and regulation. Read original evidence and track pricing and model changes.
Several user reports describe a new Codex workflow focused on image agents, including a sidebar preview, image-only conversation view, and multi-image batch editing. The feature has not yet been confirmed through an official announcement.
MiniMax has released H3 as a general-purpose omnimodal generation system for text, images, video, and audio. The official Hugging Face page is now available, although earlier reports said the weights would arrive within days.
Alibaba Cloud is offering Qwen3.8-Max-Preview through its Qwen and Model Studio services, with reasoning, vision understanding, text generation, and a large context window. Official materials still label it as a preview rather than a final release.
MiniMax has released initial weights for its H3 multimodal video model. It appears particularly strong in video editing, while its license and incomplete component release limit the scope of the open-source claim.
Several sources indicate that Qwen3.8-Max is available for testing and targets coding, research, and long-horizon tasks. Alibaba’s official material describes it as the strongest model in the Qwen family so far and says open weights are planned.
Designer Oana shared a wedding-invitation experiment made with Cola. She supplied her own illustration and used AI to help develop the lettering and layout.
An entrepreneur reports a sharp increase in organic traffic from Egypt but no completed orders, suspecting that Stripe does not directly support Egyptian merchants. The available evidence indicates a possible regional-availability issue, but does not prove that every payment route is unavailable.
Reports suggest that Tencent, Alibaba and ByteDance are consolidating AI products and enterprise services as competition shifts toward control of the enterprise agent入口.
MiniMax has released MiniMax-H3 weights on Hugging Face, extending the model beyond its hosted API and Hailuo product availability.
MiniMax H3 is now supported in SGLang Diffusion from launch day, enabling local deployment of an open-weights video model with multimodal inputs, native stereo audio, and output up to 2K at 24 fps.
MiniMax H3 is presented as an open model designed to handle multiple tasks and modalities. Reports describe text, image, audio, and video capabilities, including sound-enabled 2K video generation of up to 15 seconds.
MiniMax H3 is publicly available as an open-weight, general-purpose omni-modal model. Justine Moore shared a video generated from a single prompt asking Jim and Dwight from The Office to discuss autonomous coding agents.
A public demonstration shows MiniMax-H3 generating video locally on a single RTX 5090. However, hardware requirements vary widely by precision, resolution, and workflow, and the complete 2K pipeline is not fully local.
MiniMax has released H3 with downloadable weights. Arena.ai places it at the top among open models in its text-to-video and image-to-video arenas, while Artificial Analysis reports different rankings by task.
MiniMax says the community demonstrated H3 running on Mac shortly after release. Phosphene 3.4.0 reportedly adds MLX support, but the evidence does not establish universal Mac compatibility or a fully open-source release.
MiniMax has announced H3, an open-weights general-purpose omni-modal generation model following Hailuo 01 and Hailuo 02.
Chinese reports and social posts describe longer video generation, multi-reference inputs, timestamped prompts, and localized edits for Seedance 2.5. The claims remain unverified because the available evidence is mostly secondary and one platform says the model has not officially launched.
Telegram was temporarily unavailable across multiple Apple App Store regions, preventing new downloads and reinstalls while existing users could continue using the app. Follow-up reports said Apple suspended it after identifying content that violated its child sexual abuse material policy, and restored it after Telegram removed the material and banned the responsible user.
Media reports and user tests indicate that Seedance 2.5 is being explored for robotics training, autonomous-driving edge cases, drone navigation, and industrial training. Its controllable generation and editing tools could reduce some data-collection costs, though validation remains essential.
A discussion of Palantir’s forward-deployed engineer (FDE) model argues that foundation models will become widely available, while embedding them into complex enterprise systems will remain the harder and more valuable task.