AI Search

Claude Fable, GLM 5.2, Kimi K2.7, SCAIL 2, Minimax M3, DiffusionGemma & more AI NEWS

Welcome to the AI Search newsletter. Here are the top highlights in AI this week.

Jun 14, 2026

Claude Fable 5 is Anthropic’s most powerful generally available model, built for long, complex tasks like coding, research, vision, and knowledge work. Anthropic also introduced Mythos 5, a more restricted version for trusted cyber and science users, but Fable 5 includes extra safeguards that sometimes route risky requests to Claude Opus 4.8 instead. Read more

Anthropic says it had to suspend access to Claude Fable 5 and Mythos 5 because of a US government directive. The government reportedly raised national security concerns about a possible jailbreak, but Anthropic argues the issue was narrow and not enough reason to recall the model for everyone. Read more

DiffusionGemma is Google’s experimental open model that generates text more like an image diffusion model. Instead of writing one token at a time, it creates chunks of text in parallel, making it much faster for local interactive tasks. Read more

MoVerse turns a single image into a navigable 3D world that can be explored in real time. It expands the image into a 360° panorama, builds a 3D Gaussian scaffold, then renders moving camera videos from it. Read more

SCAIL 2 is an AI system for animating characters directly from videos without needing skeletons or pose maps. This makes it better at complex motion, character replacement, and multi-character scenes where older methods can get confused. Read more

Flex4DHuman turns one video or a few camera views of a moving person into many synchronized views. This can help create dynamic 4D human reconstructions for AR, VR, games, simulations, and virtual production. Read more

GLM is Z.AI’s flagship open-source model built for long, difficult engineering tasks like coding, debugging, and autonomous agent work. It delivers top-tier performance while being dramatically cheaper and faster than many leading closed models. Try it today

Kimi K2.7 Code is Moonshot AI’s new coding-focused agent built for developer workflows. It works through the terminal, IDEs, and CLI tools to help with programming tasks faster and more reliably. Read more

Arbor is an autonomous research agent that tests ideas, runs experiments, and learns from what works. Instead of trying random one-off solutions, it builds a tree of hypotheses and keeps improving over time. Read more

MiniMax Sparse Attention is a new attention system built to make huge AI models faster with long context. It lets the model focus on the most important parts instead of wasting compute on everything equally. Read more

Typeless is an intelligent AI voice dictation tool designed to turn your speech into polished, well-structured text in real time. It automatically cleans up your speech by removing filler words like "um," fixing mid-sentence corrections, and formatting lists across various apps and devices. Try it for free today!

Nex N2 is an AI model family built for action-heavy tasks like coding, search, and tool use. Its main trick is adaptive thinking, where it decides how much reasoning to use depending on how hard the task is. Read more

Oscar is a world model for robotics that helps robots predict what will happen after they take an action. In simple terms, it gives robots a better imagination for planning before they move. Read more

MeshFlow by Meta is a fast AI system for generating artistic 3D meshes in about one second. Instead of generating mesh pieces one by one, it uses a continuous latent space and parallel diffusion to make clean 3D geometry much faster. Read more