Most AI labs launch a model by claiming it beats everyone. Mira Murati’s lab did the opposite. Thinking Machines released Inkling on Wednesday, its first in-house model, while stating plainly in its own materials that Inkling is “not the strongest overall model available today, open or closed.” That admission is the strategy, not an apology.
Murati, who left OpenAI as chief technology officer in September 2024, is wagering that enterprises care less about the smartest general model than one they can make their own. So Inkling ships open-weight, meaning any developer can download it, inspect it, and reshape it without routing a single query through Thinking Machines’ servers.
Inkling is a mixture-of-experts system carrying 975 billion total parameters, though it activates only about 41 billion for any given task, which keeps it fast and cheap. It trained on 45 trillion tokens of text, images, audio, and video, reasoning natively across all four. Outputs stay text-only for now, including code and structured data, and it handles a context window reaching 1 million tokens.
Thinking Machines worked with Bridgewater Associates, training an open model further on the hedge fund’s financial expertise. It scored 84.7% on financial reasoning tests, beating top proprietary models at roughly a fourteenth of the cost, though that evaluation came from the two companies rather than an independent party.
The lab has raised $2 billion at a $12 billion valuation in 2025, before shipping anything at all, with NVIDIA among the backers. Investors bought Murati and her ex-OpenAI colleagues John Schulman and Lilian Weng, not customer traction. Even then, the company is modest in its approach. According to their announcement statement:
Our mission is to build AI that extends human will and judgment. We have developed a platform that lets anyone customize models, previewed an AI system built for interactive collaboration, and published novel research. Today we are advancing our mission by releasing a model we trained from scratch with the full weights available, so that people can make it their own.
Inkling sits on Hugging Face, downloadable and free of licensing fees, so local teams can run and customize it without paying subscription costs in dollars.
