Thinking Machines Lab, the AI startup founded by former OpenAI CTO Mira Murati, released its first in-house model, Inkling, an open-weight system that developers and organizations can download and modify directly. The mixture-of-experts model contains 975 billion total parameters, drawing on roughly 41 billion for any given task, and was trained on 45 trillion tokens spanning text, image, audio, and video, though its current outputs remain limited to text.
The release reflects Thinking Machines’ central bet that AI organizations can customize for themselves will outperform general-purpose models sold by major labs. Rather than positioning Inkling as a finished product, the company is marketing it as a starting point for fine-tuning through Tinker, its model-customization platform. Inkling is designed to flag uncertainty rather than guess and allows users to adjust computing effort for speed.
The company highlighted a project with Bridgewater Associates, in which a fine-tuned open-source model reportedly outperformed proprietary models on financial reasoning tests at a fraction of the cost. Thinking Machines said it built Inkling in about nine months, compared with roughly three to five years for rivals. The company now employs approximately 200 people following earlier departures, including two co-founders who left for OpenAI.
