Mira Murati’s Thinking Machines Lab Drops ‘Inkling’ : The Largest Open AI Model Ever Released in the US
In a landmark move for the open-source community, Thinking Machines Lab, founded by former OpenAI CTO Mira Murati, has officially released Inkling. This 975-billion-parameter powerhouse is now the largest open-weights AI model ever released in the United States.
The Power of Inkling
Released under the Apache 2.0 license, Inkling utilizes a sophisticated mixture-of-experts (MoE) architecture with 256 routed experts. It was trained on a staggering 45 trillion tokens and supports a massive one-million-token context window. For those with fewer resources, the lab also launched Inkling-Small, featuring 276B parameters with only 12B active at any time. This release is a major milestone in latest technology news, providing developers with unprecedented access to frontier-level intelligence.
EU Forces Android Integration
The European Union has issued binding Digital Markets Act (DMA) orders requiring Google to grant rival AI assistants, such as ChatGPT and Claude, system-level access to Android. This privilege was previously reserved for Google’s own Gemini, signaling a massive shift in the mobile ecosystem, similar to recent updates regarding Android 16 features.
Security and Robotics
In breaking news today, OpenAI’s GPT-Red model has set a new benchmark for AI safety. In automated red-teaming tests, it achieved an 84% success rate in identifying prompt injection vulnerabilities in GPT-5.1, far outperforming human experts who scored only 13%.
Meanwhile, Nvidia has launched Cosmos 3, a new Edge AI model designed specifically for robotics and vision. This launch coincides with Nvidia’s expanded "physical AI" push in Japan, aiming to integrate intelligence directly into industrial and consumer hardware.
As the AI landscape evolves, the shift toward open-weights models and increased regulation continues to redefine how we interact with technology.
