Mira Murati Launches Inkling AI Model

Former OpenAI executive Mira Murati has officially launched her new startup, Thinking Machines Lab, and its debut AI model, Inkling. Released on July 15, Inkling is an open-weight model designed to provide high-reasoning, agentic capabilities while allowing users to calibrate its 'thinking' effort. The model utilizes a mixture-of-experts (MoE) architecture, which researchers claim allows it to run more efficiently by using only 41 billion active parameters from its total 975 billion count.

Thinking Machines Lab distinguishes its approach by offering Inkling under an open-source license, facilitating direct modification by developers. This strategy contrasts sharply with the closed-model ecosystems maintained by major players like OpenAI and Anthropic. Positioned as an enterprise-grade tool, Inkling is integrated with the Tinker platform, which allows companies to fine-tune the model for specific, complex workflows. The startup aims to capture revenue through this hosting and customization ecosystem rather than metered access to the base model.

Developed with support from Nvidia’s GB300 NVL72 computing systems, Inkling was trained on a massive dataset of 45 trillion tokens spanning text, images, audio, and video. While the company acknowledges it is not currently the most powerful model on the market, it emphasizes versatility and customization for real-world enterprise tasks. The startup now employs approximately 200 staff members, marking a significant growth phase for the company following recent leadership changes in the broader AI sector.