Wan2.6, Gemini 3 Flash, GPT 5.2 Codex, MiMo V2, HY World, Seedance 1.5 Pro & more AI NEWS
Welcome to the AI Search newsletter. Here are the top highlights in AI this week.
Gemini 3 Flash is Google's latest AI model focused on speed and intelligence. It can understand and process complex information, including images and videos, and is available for developers and regular users to help with tasks like coding, planning, and learning. Read more
GPT-5.2-Codex is an AI built specifically to handle complex coding tasks like a real developer. It’s much better at big code changes, long projects, Windows workflows, and even cybersecurity work. It now leads top benchmarks that test how well AI agents work in real coding environments. Read more
HY World 1.5 is a new AI system that lets you explore and interact with 3D worlds in real time, just like playing a video game. It uses advanced memory and control systems to keep everything consistent and realistic as you move around. Read more
Do you prefer to watch instead of read? Check out this video covering the top AI news this week:
MiMo V2 Flash is Xiaomi’s new open-weight AI model that’s fast, affordable, and great at handling tough reasoning, coding, and everyday tasks. It runs up to 150 tokens per second and is designed for both general use and specialized AI applications. Read more
Resemble AI’s Chatterbox Turbo is a fast, open-source voice model that can add realistic sounds like sighs and gasps to cloned voices. It’s built for expressive, lifelike AI voices and supports text-based tags to control vocal reactions. Read more
Bytedance releases Seedance 1.5 Pro, a powerful video model that can follow complex instructions to generate diverse voices and spatial sound effects matching visuals. It supports lip-sync, dynamic camera movement, and cinematic detail for immersive video creation. Read more
GPT Image 1.5 is a major update to the image generation model, now enabling precise edits while preserving details, improved text rendering, and generating images up to 4x faster. It ranks #1 in Text to Image and Image Editing in the Artificial Analysis Image Arena.
LongCat Video Avatar is an AI that creates realistic, expressive video avatars from audio, making them move and talk naturally. It supports long, lip-synced videos and can keep the character’s identity consistent throughout. Read more
Qwen Image Layered is a cutting-edge AI that breaks down images into separate, editable layers. You can edit, move, or remove parts of an image without affecting the rest, making creative work much easier. Read more
EgoEdit is a new AI tool from Snap Research that lets users edit videos from a first-person perspective. It can make changes to a video based on simple text instructions, making video editing easier and faster for anyone. Read more
With Monica, you can use the top AI models, image generators, and video generators, all in one integrated platform. Use code AISEARCH10 to get 25% OFF ‘Unlimited Annual Plan’ within 24h of registration, or enjoy 10% OFF. Try it for free today!
FunctionGemma is a small but powerful AI model that’s great at turning everyday language into actions that apps can use. It’s designed to help build fast, private, and local agents that can call APIs based on what you ask them to do. Read more
Alibaba releases Wan2.6, a multimodal model for video and image generation capable of creating up to 15 seconds of 1080p HD narrative video with synced audio. It features reference-based character casting, multi-speaker dialogue with lip-sync, and intelligent multi-shot storytelling from simple prompts. Read more
Meta introduces SAM Audio, a unified model that isolates any sound from complex audio using text, visual, or span prompts. Its span prompting feature allows users to select a specific point in time for precise audio separation. Read more