Fine-tune a language model through an API.
Thinking Machines Lab Funding History
Thinking Machines Lab is an AI research and product company founded by former OpenAI CTO Mira Murati. Its work centers on collaborative general intelligence, with a product and model program that includes Tinker and the Inkling open-weight model family.
Company Overview
Founded 2025 · San Francisco, California
Official WebsiteThinking Machines Lab Funding Timeline
Seed
$2B · $12B valuation
The company had not disclosed a product at the time of this financing.
Led by Andreessen Horowitz. Participation included NVIDIA, AMD, Accel, ServiceNow, Cisco, Jane Street.
Source · techcrunch.com ↗
Thinking Machines Lab Founding Team
Mira Murati founded Thinking Machines Lab in February 2025. The initial team drew heavily on researchers and engineers who had worked together at OpenAI. The founding team included Lilian Weng, Barret Zoph, Andrew Tulloch, and Luke Metz.
- Mira Murati — CEO. Previously CTO of OpenAI.
- John Schulman joined the founding team as chief scientist after co-founding OpenAI and working briefly at Anthropic.
Thinking Machines Lab Products
Thinking Machines Lab Company Development
A Training Interface for Outside Researchers
Tinker gave Thinking Machines Lab a public-facing product alongside its research. The API lets users specify their training work while the service handles execution, separating the experiment design from the infrastructure needed to run it.
Thinking Machines Lab Competitive Landscape
Training Access Rather Than a Fixed Model Endpoint
Tinker's entry point is a training run the user controls. An inference-only endpoint instead serves a model already chosen and prepared by the provider. Together AI spans inference and fine-tuning, so the comparison turns on how much of the training loop a team needs to specify.
FAQ
Why is Thinking Machines' US$2B seed the largest first round in venture history?
Because it priced pedigree before product: US$2B at a US$12B post-money — roughly 4x any previous seed — for a five-month-old lab with nothing shipped, built around Mira Murati, John Schulman and other OpenAI alumni. The January 2026 collapse of the US$50B follow-on is the control experiment: when cohesion wobbled, expectations reset toward US$12B, showing how much of the price was the people.
What is a national government doing on a seed cap table?
Flying a flag: Albania invested ~US$10M — about 0.08% at the US$12B valuation — a stake small enough to require a budget amendment in Tirana. It previews the sovereign-money pattern across the set: the UK's Sovereign AI Fund in Ineffable's seed, Temasek in AMI Labs', Saudi HUMAIN's US$3B in xAI's Series E — states buying frontier-lab exposure as policy, at symbolically sized tickets.
Why did the planned US$50B follow-on collapse?
Because the product lagged the price: by January 2026 the NYT reported co-founders pressing for a sale or merger over direction and release pace, followed by departures including CTO Barret Zoph. Expectations reset toward ~US$12B — a year of churn without a step-up. The counter-move is Inkling, the July 2026 open-weight family (975B and 41B parameters), shipped ahead of the reported US$5–6B raise.
How does Thinking Machines differ from Sutskever's SSI, given both are ex-OpenAI bets?
Opposite structures: SSI's charter bans commercialization outright, while Thinking Machines shipped Tinker in late 2025 and the open-weight Inkling family by July 2026, leaning on operational revenue after the US$50B round fell apart. SSI's US$32B came with no product plan; TML's US$12B seed demands one — the pending round will price whether OpenAI pedigree alone still clears frontier multiples.
Why does the seed syndicate read like a hardware coalition?
Because chipmakers seed future GPU buyers: NVIDIA, AMD, Cisco and ServiceNow joined a16z and Jane Street — the same strategic pattern as SSI's round and AMI Labs' five-co-lead group. The hedge cuts both ways: after the US$50B collapse, NVIDIA's reported ~US$2.5B consideration in the pending round shows hardware capital keeping its option open on a lab whose first seed it already priced.
How can I distinguish research independence from the interests of hardware investors?
A hardware company's participation signals a financial relationship, not automatic control over the lab's research agenda or compute suppliers. For Thinking Machines, read ownership, governance and commercial arrangements separately. The investor list alone cannot establish exclusivity, board control or product commitments.