One of the most closely watched startups in artificial intelligence has finally shipped a product. Thinking Machines Lab, founded by Mira Murati after she left the role of chief technology officer at OpenAI, released its first model, an open system called Inkling, TechCrunch reported. Notably, the company is not claiming to have built the best model, but a different kind of product.
What was released
Inkling is what the industry calls an "open-weight" model: the underlying model is made available for others to download, run and adapt, rather than being locked behind a company's own service. It is a large "mixture-of-experts" system, a design that holds a very large number of parameters, 975 billion in total, but activates only a fraction, about 41 billion, for any given task, which saves on computing cost, and it was trained on 45 trillion units of data spanning text, images, audio and video, according to TechCrunch.
The striking part is the company's own framing. Thinking Machines said plainly that Inkling is "not the strongest model available today, closed or open," per reporting by Fortune, emphasizing instead efficiency and adaptability. That is an unusual pitch in an industry that markets each release as a new frontier.
The strategy: adapt, don't dictate
The reasoning behind that pitch is a real business bet. Where OpenAI, Anthropic and Google sell access to closed, general-purpose systems, Thinking Machines argues that organizations able to adapt a model to their own needs will get more out of it than they would from a one-size-fits-all product, as TechCrunch summarized the company's thesis. The company pairs the open model with its own tools for fine-tuning, the process of adjusting a model on a customer's data for a specific job, betting that businesses will pay for that customization and support.
It is a similar philosophy to Meta's Llama and to China's DeepSeek, both of which have won users by opening up their models. The wager is that in an enterprise market, control and tailoring can matter more than topping a leaderboard.
The company behind it
Thinking Machines has been a sensation on funding alone. Murati, a prominent figure from OpenAI, drew a wave of senior AI researchers to the venture. The startup raised about $2 billion in a seed round that valued it at roughly $12 billion in July 2025, before it had a product, one of the largest such rounds on record, according to TechCrunch. It has since employed around 200 people.
The money has also brought scrutiny. A far larger follow-on financing, reported at around $50 billion, was said to be coming together late last year but had stalled by January, as backers sought more evidence of a product before committing at such a price, per TechCrunch's account. Seen that way, Inkling is not just a technical release but a demonstration, the first hard evidence that a company valued in the billions on promise alone can actually ship.
Why it matters
For the AI market, Inkling adds weight to the open-model camp at a moment when the biggest labs are doubling down on closed systems and preparing for the public markets. If Thinking Machines is right that customization beats raw capability for many business buyers, it could carve out a durable niche without ever winning the headline race on benchmarks. If it is wrong, a $12 billion valuation will look hard to justify. Either way, the company has finally moved from the most expensive idea in AI to a product people can actually run, and the market can now judge it on more than its founder's reputation.
