Mira Murati's AI lab steps out of the shadows with an open-weights model built for customization, backed by $2B and a who's who of tech investors
Mira Murati spent roughly 18 months after leaving OpenAI building something quietly. No product launches, no demos, no press tours. Just infrastructure work, a massive fundraise, and a lot of hiring. Now the curtain is coming back, just a little, with the debut of Inkling, the first public model from Thinking Machines Lab.
Inkling is an open-weights model, meaning developers can download, modify, and deploy it without routing everything through a centralized API.
What Thinking Machines actually built #
The lab was founded in February 2025, and it moved fast. Within weeks of launching, Thinking Machines closed what research suggests was the largest seed round in venture capital history: $2B at a $12B valuation.
Backers include Andreessen Horowitz, Nvidia, AMD, Cisco, and Jane Street. Nvidia also signed a multiyear chip supply agreement in March 2026, locking in access to its Vera Rubin accelerators.
The company operates as a public benefit corporation with around 100 employees.
Murati herself was OpenAI’s CTO and served as interim CEO during the board’s brief and chaotic attempt to oust Sam Altman in late 2023. She resigned from OpenAI in September 2024 to start Thinking Machines.
Why open-weights matters right now #
Inkling enters the open versus closed model conversation as a model explicitly designed for customization. The framing from the lab emphasizes human-AI collaboration and strength in scientific and programming tasks, which points toward developer and enterprise use cases rather than consumer chatbot territory.
Open-weights deployment changes the economic calculus for anyone building on top of AI. Instead of paying per API call to OpenAI or Anthropic, a developer can run Inkling on their own infrastructure, fine-tune it for their specific use case, and keep the resulting model private.
The AI infrastructure investment wave and what it means for markets #
Nvidia’s involvement here is notable beyond the chip supply deal. When Nvidia takes equity in an AI lab and also commits to supply its next-generation accelerators, it’s effectively making a bet that this lab becomes a significant buyer. Vera Rubin accelerators, announced for 2026, represent Nvidia’s next architectural leap after Blackwell.
AMD has been aggressively positioning its MI300 series as a cost-competitive alternative to Nvidia hardware. An AI lab backed by both chip giants has optionality on the hardware side that most competitors don’t.
There is one thing worth flagging for crypto-adjacent readers. A low-cap token trading under the ticker TML and using the Thinking Machines Lab name exists in secondary markets. The lab has no verified connection to that token.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our