Mira Murati's AI Startup Just Released Its First Model. It Wants Companies to Own Their AI, Not Rent It.

Key points

• Thinking Machines Lab released Inkling on Wednesday, its first publicly available AI model, roughly nine months after the company began building.
• Inkling has 975 billion total parameters (a measure of model complexity) but uses only about 41 billion for any single task, keeping it faster and cheaper to run than its size suggests.
• The model is open-weight, meaning any business or developer can download and modify it for free, unlike ChatGPT or Claude.
• A joint study published in late June found that a custom model built using similar open-weight methods scored 84.7% on financial reasoning tests at roughly one-fourteenth the running cost of top proprietary AI models.
• Thinking Machines plans to make money through Tinker, its paid platform for training and adapting the model, not from the model itself.

Mira Murati spent years as the chief technology officer at OpenAI. Now she is betting against the thing OpenAI sells.

Thinking Machines Lab, the startup Murati founded after leaving OpenAI, released its first AI model on Wednesday. It is called Inkling, and it works differently from the headline AI products most people have heard of.

ChatGPT, Claude, and Google Gemini are all closed systems. You pay to use them. The company that built them keeps full control. Inkling is open-weight, which means any business or developer can download the full model and reshape it to fit their own needs, at no charge from Thinking Machines.

That is the whole pitch. The company believes AI that a business trains on its own data and expertise will outperform a general-purpose model that a big lab sells to everyone.

The evidence backing that argument is genuinely interesting. Bridgewater Associates, the world's largest hedge fund, worked with researchers to take an existing open-source AI model and train it further on Bridgewater's own financial knowledge. That custom model scored 84.7% on financial reasoning tests, beating top proprietary AI products, while costing roughly one-fourteenth as much to run. The results come from a study the two companies published jointly in late June, not from an independent reviewer, so treat that with appropriate scepticism.

Microsoft chief executive Satya Nadella made a related point in a blog post on Sunday. Companies using closed AI products, he wrote, pay twice: once in subscription fees, and again by handing over the business knowledge baked into all their prompts and corrections, knowledge that can feed back into future versions of a model they do not own.

So is Inkling actually any good?

Thinking Machines says plainly that Inkling is not the strongest AI model available today, closed or open. What it aims for instead is steady, well-rounded performance at lower cost. The company says Inkling uses one-third as many processing steps as a competing model from Nvidia to reach the same score on coding tests. It can also flag when it is uncertain rather than guessing, and users can dial its thinking effort up or down depending on whether they want a careful answer or a quick one.

Inkling is enormous under the hood. It was trained on 45 trillion tokens, which are the chunks of text, image, audio, and video data that AI systems learn from. But because of its architecture (a mixture-of-experts design, where only a slice of the model activates for each task) it stays practical to run.

Thinking Machines has been careful about one thing: Inkling was partly trained using outputs from other open-weight AI models, a common practice called distillation. The company says its next model will drop that approach entirely.

Where does Thinking Machines make money? Not from the model download. Revenue comes from Tinker, its paid platform for helping businesses fine-tune and host their own custom versions of Inkling.

That is the honest takeaway here. If your organisation has real, specific expertise locked inside your staff's heads and your past work, a custom AI model trained on that knowledge may outperform a general one at a fraction of the cost. The path to that is now a bit shorter.