Mira Murati Launches Inkling, Thinking Machines' First Open Model, and Admits It's Not the Strongest on the Market
Thinking Machines, Mira Murati's startup valued at $12 billion, launched Inkling on July 15, 2026, a 975 billion parameter open model the company itself says isn't the market's strongest.
Who Thinking Machines is and what Inkling is
Thinking Machines Lab, the startup founded by former OpenAI CTO Mira Murati and valued at $12 billion, released its first broadly usable AI model, called Inkling, on July 15, 2026, according to reporting from TechCrunch, Fortune and Axios. Unlike the flagship models from OpenAI, Anthropic or Google, Inkling is open weight: any developer or company can download the model and modify it directly, instead of accessing it only through a closed API.
Technical specifications
Inkling runs on a mixture of experts architecture, with 975 billion total parameters, though it only draws on about 41 billion of them for any given task, according to TechCrunch. The model was trained on 45 trillion tokens of text, image, audio and video, and processes all four modalities natively at once, without needing intermediate conversions between formats.
Calibrated answers and a reasoning effort dial
One of the model's differentiators is its emphasis on calibrated answers: Inkling was designed to flag uncertainty rather than simply guess when it lacks enough confidence, according to TechCrunch. The system also lets users dial reasoning effort up or down, trading speed for depth of analysis depending on each task's needs. In one benchmark cited by the company, Inkling used a third as many tokens as Nvidia's Nemotron 3 Ultra to reach the same coding performance.
A model the company itself says isn't the best
Thinking Machines explicitly states that Inkling is not the strongest model available today, open or closed, according to coverage from Startup Fortune. Rather than competing directly on raw performance, the company positions Inkling as a customizable foundation for enterprises that want to tailor a model to their own needs, instead of just consuming a generic API. The company, for its part, is not looking to monetize the model itself yet: its current revenue comes from Tinker, a model fine tuning tool sold to customers such as the hedge fund Bridgewater Associates.
Why it matters for those building products on top of AI models
For agencies and tech companies building products on top of third party models, Inkling widens the menu of open weight options available to anyone who wants to train or fine tune a model on their own data, without depending only on a closed API. Thinking Machines' honesty about the model's limitations is also a relevant data point: not every open weight release needs to compete for the top of the performance ranking to be useful, especially when the goal is to serve as a customizable foundation rather than a finished, ready to use product.