Microsoft just launched MAI-Image-2.6, debuting at number two on the Arena text-to-image leaderboard ahead of models from Google, Meta, and xAI. The new release gains +79 Elo overall over MAI-Image-2.5, with text rendering alone jumping by +91 Elo alongside noticeably cleaner portraits and commercial graphics. It puts Microsoft firmly in the top tier of frontier image generation.
Tencent just introduced WorldClaw, an agentic system that turns a single text prompt into an explorable, editable 3D open world. It maps large-scale terrain, populates regions with textured meshes, and deploys render agents to tweak contacts and object poses—all while keeping every single asset reusable.
Researchers at Switzerland's Paul Scherrer Institute just built XtalPaint, an open-source model that reconstructs missing atoms in crystal lattices with a 97% success rate. Adapted from Microsoft's MatterGen, it treats crystal gaps like erased parts of an image, applying diffusion noise only to unknown spots while using known atoms as anchors. Why it matters: X-rays often miss light elements like hydrogen, locking thousands of promising compounds out of computer simulations; this tool could unlock them for battery and clean-energy research.
A mysterious "mona-lisa-1" model just surfaced on LM Arena, fueling speculation that OpenAI is testing a new GPT image generation model. Community reports note the outputs contain SynthID watermarks and visibly outperform Image 2, though OpenAI has yet to officially confirm the model.
xAI just rolled out Imagine Image 2.0 as the new Quality Mode across grok.com and its mobile apps. Built for practical design work, the model brings precise local tweaks via a magic wand tool, background removal, and multi-reference support for up to 5 images. It also jumped to second place globally on Arena's text-to-image and editing leaderboards.
Multi-Agent CAD (MAC) v2 turns natural-language prompts into editable, printable 3D engineering parts and assemblies. By decoupling planning, coding, and visual verification across multiple agents, it cuts token usage by up to 116x and costs by 13x over baselines. The open-source tool exports STEP, STL, and simulation-ready URDF files.
JD just open-sourced JoyAI-Video-Edit, an autoregressive diffusion model that edits video streams on the fly via text instructions. Powered by a 16B-parameter MMDiT, it handles clothing swaps, object removal, and style edits at 30 FPS in 720p, and can run at 24 FPS on a single RTX 5090.
Tencent just introduced Hunyuan3D-Buffalo 1.0, a unified multimodal framework handling 3D generation, visual reasoning, natural language editing, and part decomposition in one pipeline. Powered by a shared VLM backbone and DiT modules, it lets creators edit or break down 3D meshes using plain text prompts without chaining multiple standalone tools.
MiniMax just unveiled H3, an omni-modal generation model that natively handles text, images, video, and stereo audio in a single context. It generates up to 15-second clips at 2K resolution at a fraction of competitors' costs. MiniMax also plans to open-source the model weights in the coming days, taking a rare open-weights bet against closed video giants.
Ideogram and Pruna AI just launched P-Image-Ideogram, a family of Pareto-optimal image generation models engineered for peak cost-efficiency. Starting at just $0.003 per image, the models support native 1K and 2K outputs across four quality tiers with 3 to 8 seconds of latency. Why it matters: High-fidelity image generation is quickly turning into an ultra-cheap utility, putting immediate pricing heat on rivals.
Google Earth just rolled out an AI tool powered by Nano Banana 2 that let users reimagine any real-world location using satellite and 3D map data. But just a day later, Google pulled the plug after generated screenshots violating safety policies surfaced online. The tool is now temporarily shelved until stronger guardrails are built.
Ideogram just launched Object Remover, a dedicated AI tool that scrubs unwanted items, text, and logos from photos. Instead of leaving smudged artifacts, it erases the surrounding shadows, reflections, and subtle lighting shifts while cleanly rebuilding the background underneath. Ideogram says it ranks first on the RemovalBench benchmark while beating rivals on per-request pricing.
Replit just launched Replit Design, a new AI-powered creative suite built to make design accessible to anyone. The tool aims to turn rough concepts into finished visuals with ease, marking a clear push by the company beyond code into full creative workflows.
Ai2 just launched the OlmoEarth Platform, an infrastructure stack built to run Earth observation foundation models at planetary scale. It can process continent-sized satellite datasets in roughly a day for fractions of a penny per square kilometer. In a test mapping North America wildfire risk, it coordinated 19,600 CPUs and 994 GPUs to deliver a 155x speedup.
ReDesign reverses flat raster images back into editable layer hierarchies with typography, vector shapes, and depth order. An agentic vision controller orchestrates tools like Qwen-Image-Layered and SAM 2 to recursively unpack elements, and the code is now open-source on GitHub.
Adobe and Johns Hopkins researchers just introduced Wonder, a video world model that turns a single image or video into an interactive environment. Users can explore scenes at 16 FPS with 6-DOF camera controls, and the model maintains consistent geometry even after roaming around for over a minute.
Google just open-sourced GNM, a generative 3D head model that gives creators granular slider-level control over facial features, down to the eyeballs, teeth, and tongue. With over 600 independent identity and expression parameters under an Apache 2.0 license, it tackles the "same-face" look and hollow-mouth flaws common in AI-generated humans.
Midjourney just acquired personalized astrology app Co-Star to kickstart its mobile app efforts. Co-Star founder Banu Guler is joining as chief design officer to help build Midjourney's first standalone apps, including a dedicated image generation tool. It marks a big shift for an AI heavyweight currently confined to Discord and the web.
Netflix’s Eyeline Labs just unveiled ID-V2V, a framework that restyles video lighting and scenes using a single edited keyframe. Crucially, it locks down the actor's facial identity, gaze, and lip sync, automatically correcting misaligned input frames. It maxes out at 720p, and with an 80GB model weight, you will need serious local hardware to run it.
Developers just open-sourced OpenDreamer, a playable Minecraft world model that reproduces Google's closed-source Dreamer 4. It acts as an action-conditioned video generator without pre-programmed game logic, and the team released the full training pipeline, weights, and a browser demo.
