Qwen3, Phi4, new image editors, battery-free AI, open-source Deep Research, Xiaomi Mimo, Deepseek Prover
Welcome to the AI Search newsletter. Here are the top highlights in AI this week.
Qwen3 is the newest large language model from the Qwen family, featuring both massive and small models that outperform previous versions and many competitors. It was trained on a huge, diverse dataset (36 trillion tokens, 119 languages), and comes in both dense and MoE (Mixture of Experts) versions, offering top performance in coding, math, and reasoning tasks. Read more
A new shortcut in AI design helps solve memory bottlenecks, making AI models faster and more efficient. By rethinking how AI stores and accesses information, researchers have reduced delays and improved performance, especially for large language models and other demanding tasks. This could lead to quicker, smarter AI that works better on everyday devices. Read more
Microsoft’s Phi small language models, like Phi-4-reasoning, bring powerful AI reasoning to smaller, faster models. These models can solve complex math and science problems, rivaling much larger AIs, and are optimized to run efficiently even on regular PCs and mobile devices. Read more
Do you prefer to watch instead of read? Check out this video covering all the highlights in AI this week:
WebThinker is a research agent that lets large reasoning models search the web and write detailed reports. Instead of relying only on what they already know, these models can now explore online information, gather data, and combine it with their reasoning to create comprehensive answers. Read more
ICEdit is a new AI tool for editing images using natural language instructions, needing very little training data or computing power. It uses a smart combination of in-context prompts and efficient tuning, achieving high-quality edits that compete with commercial models like Gemini and GPT-4O, while being open-source and fast. Read more
We’re partnering with Dell to give away a Dell Precision 5690 Workstation with a RTX 5000 Ada. This is a powerful yet portable laptop, with a built-in NPU that’s optimized for AI. Only available to USA or Canada residents. Enter for FREE here
MiMo is an open-source project by Xiaomi for AI-powered image and video editing. It lets users apply advanced edits and effects to their photos and videos using easy-to-use tools and AI models. Read more
HiDream-E1 is an open-source image editing AI that builds on the HiDream-I1 model. It’s designed for instruction-based editing, letting users change images by describing what they want in natural language, and is available for anyone to try or build on. Read more
DeepSeek-Prover-V2 is an open-source AI model focused on solving and explaining math and logic problems. It can generate step-by-step proofs and explanations, making it useful for learning or checking complex math work. Read more
F-Lite is a lightweight, open-source AI model for fast and efficient image generation and editing. It’s designed to run on devices with limited resources, making advanced image AI accessible to more people. Read more
Mellum is a purpose-built large language model for developers, now open-source and available on Hugging Face. It’s designed to help with coding tasks, code generation, and developer workflows, aiming to boost productivity for programmers. 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!
Google’s NotebookLM now lets you listen to AI-generated audio summaries of your notes in over 50 languages. The Audio Overviews feature turns your uploaded documents into podcast-like conversations between two AI hosts, making it easier to understand and remember information, especially for people who prefer listening over reading. This update breaks language barriers and makes learning more accessible worldwide. Read more
Researchers have developed a new standard for battery-free, AI-enabled devices that can run advanced AI tasks without needing to recharge. These devices use energy from their environment (like light or movement) and can process information on their own, which could lead to smarter, longer-lasting gadgets and sensors. This breakthrough could make AI-powered tech more sustainable and widely available. Read more