MiroMind, the AI company founded by Chinese billionaire Chen Tianqiao, will suspend its MiroThinker service in China, Hong Kong, and Macau from May 12 after regulators imposed a firewall between its US and China operations following the Manus acquisition block.
Anthropic's new dreaming feature lets Claude Managed Agents review their own sessions, extract patterns, and improve over time while using less memory. Developers can control whether updates are automatic or reviewed by humans.
In court, OpenAI president Greg Brockman revealed that Elon Musk demanded 51% control of the startup to fund a Mars city — and stormed out when they refused.
In court, OpenAI president Greg Brockman revealed that Elon Musk demanded 51% control of the startup to fund a Mars city — and stormed out when they refused.
Mistral AI released Mistral Medium 3.5, a 128B parameter open model that outperforms Claude Sonnet 4.5 on agent benchmarks, along with cloud updates to Mistral Vibe and Le Chat.
Qwen open-sourced FlashQLA, a high-performance operator library for linear attention, claiming up to 3x forward speedup on H200 over FLA Triton kernel and up to 5.33x over FlashInfer in multi-GPU setups.
ByteDance, Zhipu AI, and Alibaba land on TIME's first-ever AI-focused list, alongside OpenAI, Anthropic, and a surprise 'no-Western-chips' label for Zhipu.
Alibaba Cloud's new Qwen-Image-2.0-Pro model combines text-to-image generation and editing into one, with improved multilingual text rendering and up to 2K resolution output.
Perplexity details its two-stage training process for web search agents, showing its open-source Qwen-based model outperforms GPT-5.4 and Claude Sonnet on accuracy while costing a fraction of the price.
Alibaba's Tongyi Lab has released Qwen3.6-27B, a 27-billion-parameter multimodal AI model under the Apache 2.0 license. It's designed to run on local hardware and claims to outperform Google's Gemma4-31B and even rival Anthropic's Claude 4.5 Opus in some benchmarks.
