Z.ai released GLM-5.3, enhancing coding capabilities with a 50% performance improvement over GLM-5.2 and achieving state-of-the-art results on multiple public benchmarks. Additionally, GLM-5.3 demonstrates advanced emergent cyber defense capabilities, excelling in vulnerability discovery and exploitation tasks on CyberGym.

x.ai released Grok 4.6, enhancing its capability with long-running agents and complex tasks across multiple steps in fields like STEM and software engineering. The model exhibits frontier intelligence, achieving high performance benchmarks and demonstrating improved self-verification in project development, especially in generating visual and interactive applications.

Meta has released Muse Glimmer, an open-weight model with 30 billion parameters optimized for always-on local agents, enabling complex multi-step tasks that involve planning, error handling, and long-horizon execution.

Alibaba released open weights for the Qwen3.8-27B multimodal dense model, featuring 27 billion parameters that excel in coding and office tasks. The model supports 262K native context, extendable to 1 million tokens, and is available under an Apache 2.0 license for developers.

Meta Superintelligence Labs launched Muse Code (beta), a terminal coding agent that leverages Muse Spark 1.2 to execute complex software engineering tasks by coordinating persistent subagents for enhanced efficiency. Muse Spark 1.2 improves code generation, debugging, and workflow integration, optimizing performance for extensive coding projects and long-horizon tasks.

xAI launched Imagine Image 2.0 in Grok's Quality Mode, providing users with features like precise edits, smart resize, and workflow templates across web, iOS, and Android platforms.

Claude now enables sessions to communicate directly by sending summaries to each other, allowing for seamless continuity without the need to re-explain tasks. This feature enhances workflow efficiency in multi-session interactions.

OpenAI introduced a new feature in ChatGPT's desktop app that allows the AI to remember user activity across applications and websites, enhancing personalization and reducing the need for repeated explanations. This Computer History feature includes a timeline view and improved privacy controls, rolling out to Pro, Business, and Enterprise users.

Claude implements a system to mark AI-generated content, enhancing the identification and management of such posts within its framework. This feature aims to address challenges associated with distinguishing between human and AI-generated materials.