Skip to content
AI.info

creators

Mira Murati: co-founder and CEO of Thinking Machines Lab

Mira Murati was OpenAI's CTO until 2024 and founded Thinking Machines Lab in 2025, which shipped Tinker in 2025 and the open-weights Inkling model in 2026.

Mira Murati: co-founder and CEO of Thinking Machines Lab

Mira Murati (born Ermira Murati, 16 December 1988, in Vlorë, Albania) is an Albanian-American executive. She attended Pearson College UWC, took a BA at Colby College in 2011 and a Bachelor of Engineering in mechanical engineering at Dartmouth College in 2012. Early in her career she was a summer analyst at Goldman Sachs and an intern at Zodiac Aerospace, then a product manager at Tesla (2013–2016) on the Model X programme, followed by product and engineering work at the augmented-reality company Leap Motion (2016–2018). Murati joined OpenAI in 2018 as vice president of applied AI and partnerships and became chief technology officer in May 2022, a post she held until she left in September 2024. She led OpenAI's work on ChatGPT, DALL·E, Codex and Sora while overseeing its research, product and safety teams, and was chief technology officer across GPT-4's release in March 2023; in November 2023 she was briefly interim chief executive for three days during the board's removal and reinstatement of Sam Altman. In February 2025 she founded Thinking Machines Lab, a San Francisco research and product company structured as a public benefit corporation; its own site names her cofounder and CEO. It opened with 29 researchers and engineers hired from OpenAI, Character.AI, Google DeepMind and other labs, and raised $2 billion in July 2025 at a $12 billion valuation in a round led by Andreessen Horowitz, with Nvidia, AMD, Cisco and the Albanian government among the investors. The lab shipped Tinker, an API for fine-tuning open-weight models, on 1 October 2025, announced a gigawatt-scale NVIDIA partnership on 10 March 2026, set out its case for "interaction models" on 11 May 2026, and released Inkling, an open-weights mixture-of-experts model with 975 billion total and 41 billion active parameters, on 15 July 2026.

Specialization
large language models, open-weights models, model fine-tuning
Country
United States

Work