AI Article Tags

Browse articles by professionally maintained tags. Every tag opens a focused collection with its own editorial introduction and closely related published articles.

Qwen-Image Shows how the Qwen-Image line branches, from its strength in text layout to the realism upgrade in 2512 and the layer decomposition of Layered, and where each applies. AI.com Explains how onboarding and permissions work on ai.com, weighing username reservation, cross-app execution and isolated encrypted execution against each other. Z-Image Introduces the Z-Image open image model line, mapping the non-distilled backbone, single-stream diffusion design and free inference quota to concrete creative and developer needs. Code Supernova Breaks down the agentic coding behaviour of Code Supernova, from long context and image input to multi-step execution, and what each changes on real cross-file rework. LTX Video Traces how the LTX Video models and the studio platform divide the work, from open video models and storyboard generation to timeline editing and voice-over. Doubao Phone Assistant Looks at ByteDance’s Doubao Phone Assistant, how it taps, swipes and types across apps once it holds system permissions, and what the nubia M153 preview device realistically covers for early users. Gemini Robotics Explains how Google DeepMind’s Gemini Robotics works as a robot’s high-level brain, how ER 2 tracks progress and recovers from errors mid-motion, and where the Atlas collaboration stands. Google Disco Breaks down Disco, Google Labs’ experimental browsing project, and its first feature GenTabs, showing how open tabs become small usable apps and what the waitlist and privacy notes mean. GLM-5 Follows Z.ai’s open-weight GLM-5 flagship line, from 744B base weights and eight-hour autonomous runs to the cheaper 320B Flash, and shows how the generation balances capability against cost. 360 Maps out the 360 AI lineup from Zhinao and the AI Browser to Zhitu, Hongtu and the Nami comic-drama pipeline, showing which tasks each line takes on, who it serves, and where the Mem360 memory layer fits. LongCat-Video-Avatar Read the LongCat-Video-Avatar audio-driven avatar video model line, and see how it handles long-clip stability, multi-person identity and natural motion. Tongyi Fun Explain what makes up Tongyi Fun’s enterprise audio stack, and how Fun-ASR recognition and Fun-CosyVoice synthesis cover meeting notes and support quality checks. HunyuanWorld Break down how HunyuanWorld-Voyager grows an explorable 3D world from a single image, and what native 3D output, world caching and custom camera paths solve. Hunyuan-MT Follow Hunyuan-MT from the 7B cloud model to the 1.8B edge model, and see which localization jobs terminology control, long context and format preservation each serve. MiniMax Music Explain how MiniMax Music turns lyrics and a style description into a finished song, and what realistic vocals, mood control and the open-source Music Skills each handle. LongCat-Image Focuses on how LongCat-Image pits a 6B DiT backbone against larger systems, and how bilingual text rendering and controllable editing behave in real design and everyday creation work. GLM-TTS Centres on GLM-TTS voice cloning from three seconds of audio and its emotion control, covering how two-stage generation with reinforcement learning lowers error rates and where narration and service use it. Firecrawl Explains how Firecrawl splits scraping, crawling, search and URL mapping, and how interactive actions, callbacks and multiple output formats feed agents and retrieval pipelines with clean web data. PaddleOCR Breaks down PaddleOCR's detection and recognition, layout parsing and 0.9B document model, tracing the path from multilingual or awkward scans to structured Markdown and JSON output. LangChain Walks through LangChain's component model and chain orchestration, then shows how prompt templates, retrieval augmentation, memory and tracing support RAG pipelines, assistants and automation in production. MiMo-V2 Focuses on the MiMo-V2 line from Xiaomi and compares Flash, Pro, Omni, TTS and V2.5 across parameter scale, context length and multimodal capability. Google Gemma Lays out the two branches of Google's open-weight family: Gemma 4 leans on reasoning and local deployment, VaultGemma on differential privacy, each carrying its own licence and compliance calculus. HunyuanOCR Breaks down how a one-billion-parameter OCR model completes detection, recognition and layout parsing in a single pass, and covers deployment for receipts, subtitles and multilingual image translation. Notion AI Sets out how Notion AI handles writing, summarising, translation and question answering across pages, wikis and databases, and maps a knowledge routine that individuals and teams can adopt as is.