Mira Murati’s New Startup to Build Custom AI Models for Businesses

Mira Murati, the former chief technology officer at OpenAI, is quietly building a new AI venture with an ambitious goal: developing custom AI models tailored to a company’s specific business needs. Though her startup, Thinking Machines Lab, has kept a low public profile, a new report from The Information offers early insight into what the company may be working on.

According to the report, Murati’s company is focused on creating AI tools designed around a client’s key performance indicators (KPIs)—metrics used to track business goals. The startup reportedly plans to use reinforcement learning, a technique where AI systems learn by repeatedly attempting a task and receiving feedback in the form of rewards or penalties. This method helps models improve over time through trial and error.

Murati’s Background and Funding Push

Murati joined OpenAI in 2016 and rose to a senior leadership role, serving briefly as interim CEO during the company’s high-profile leadership shakeup in 2023, when Sam Altman was briefly ousted and then reinstated. She left OpenAI in September 2024 and has since been raising funds and recruiting talent for her new venture.

Thinking Machines Lab has reportedly raised $2 billion at a $10 billion valuation, with much of that momentum driven by Murati’s reputation and network in the AI industry.

Building Models Faster Using Open-Source AI

Training advanced AI models from scratch can be time-consuming and expensive, so Murati is said to be pursuing a more modular approach. Rather than building full models from the ground up, the startup aims to “pluck” specific layers from existing open-source AI models and recombine them to suit specific tasks. Because open-source models don’t require an API to access, this strategy could allow the company to move quickly and operate more independently than larger competitors.

To understand this, consider how a model like ChatGPT processes information. Input data moves through a stack of neural network “layers,” each handling a different aspect of language understanding and response generation. By selectively combining certain layers, Murati hopes to construct powerful AI models tailored to business use cases—faster and more efficiently than traditional methods.

It remains unclear which tasks Thinking Machines Lab’s models will target, or what the cost of developing a custom solution will be.

A Consumer Product May Also Be in the Works

Although much of the focus appears to be on enterprise AI tools, The Information also reports that Thinking Machines Lab is developing a consumer-facing product, potentially a rival to ChatGPT. Details about its features or timeline remain unclear.

Ties—and Distance—from Meta

Murati has reportedly been among several AI founders in recent talks with Meta CEO Mark Zuckerberg, who is investing heavily in AI through talent acquisition and startup partnerships. However, the report notes that discussions between Murati and Zuckerberg “didn’t get far.”

Looking Ahead

While Thinking Machines Lab remains largely under wraps, Murati’s strategy points to a future where AI tools are not one-size-fits-all, but custom-built to meet the exact goals of each business. If her team can successfully combine open-source components into specialized, reinforcement-trained models, it could offer a faster, leaner path to enterprise AI adoption—and challenge the dominance of today’s biggest model providers.

Murati’s next steps may help define whether modular AI can deliver not just speed, but real competitive edge.

Editor’s Note: This article was created by Alicia Shapiro, CMO of AiNews.com, with writing, image, and idea-generation support from ChatGPT, an AI assistant. However, the final perspective and editorial choices are solely Alicia Shapiro’s. Special thanks to ChatGPT for assistance with research and editorial support in crafting this article.