Thinking Machines Lab, an AI startup founded by Mira Murati, is betting on the future of customizable open-weight AI models. The company is building multimodal AI systems that are designed around human-AI collaboration. A key tenet of the company’s vision is that AI should be distributed widely and equitably.

Mira Murati, the founder and CEO of Thinking Machines, is the former Chief Technology Officer of OpenAI. She also briefly served as its interim CEO when Sam Altman was ousted from OpenAI for a few days. Ms. Murati joined OpenAI in 2018 and quick climbed up the ranks. In September 2024, she decided to step away from OpenAI and instead take on the challenge of building her own AI company.

Thinking Machines Lab was launched in February 2025. She recruited several OpenAI researchers to join her, although some of those individuals later decided to move back to OpenAI. In April 2025, just a couple months after launching, the startup raised $2 billion in seed round capital from investors. Thinking Machines is structured as a public benefit corporation (PBC), similar to OpenAI and Anthropic. Public benefit corporations are for-profit entities that are required to balance the advancement of social and environment goals with generating financial returns for shareholders.

The AI startup launched its Tinker platform in October 2025, an application programming interface (API) for fine-tuning open-weight and open-source models. Open-weight models only provide users with access to the final trained numerical parameters (referred to as the “weights”), but generally do not disclose the source code or training datasets. In contrast, open-source models provide full transparency into the source code.

In July 2026, Thinking Machines released an open-weight AI model called Inkling. It is a broad-based, generalist model that can adapt to many different types of workflows. Inkling is a “mixture-of-experts” system that was trained on 45 trillion tokens of text, images, audio, and video. The startup also released a product called Inkling Small, which achieves strong results at a lower cost than Inkling. User customization is a key component of the Inkling models.

Thinking Machines has scored a number of high-profile partnerships. In March 2026, Thinking Machines entered into a multi-year partnership agreement with NVIDIA. Pursuant to the agreement, Thinking Machines has committed to deploy at least one gigawatt of NVIDIA’s Vera Rubin platform.

“NVIDIA’s technology is the foundation on which the entire field is built,” stated Thinking Machines CEO Mira Murati. “This partnership accelerates our capacity to build AI that people can shape and make their own, as it shapes human potential in turn.”

In September 2026, Thinking Machines entered into a strategic partnership with Crusoe to power inference for Thinking Machines’ AI models. Crusoe is a vertically integrated AI infrastructure that is building data centers for OpenAI, Google, and others.

As competition for elite AI researchers intensifies, AI companies have been poaching top talent from one another in the heated AI race. While Thinking Machines is on a positive trajectory, it remains to be seen whether the AI startup can maintain its upward momentum.