OpenAI’s former CTO just dropped a very different kind of AI model

Mira Murati’s Thinking Machines released its first model, Inkling. It might be perfect for your business.

When Mira Murati left the chief technology officer post at OpenAI in 2024, it caused quite the stir. Murati had been a figurehead for the market-leading AI brand, and because she departed when the controversial firm was on the ascendant, people were curious about her plans. Her next venture, Thinking Machines, kept its intentions pretty quiet. When she raised capital and poached engineers from her former employer, Murati hit the headlines again, but there were still few details available about what she was working on. 

Now, as it has released its first AI model, Inkling, we know exactly what Thinking Machines has been doing. And it’s fascinating. 

As much of the rest of the AI industry has zigged with its model designs, Murati’s firm has zagged—the model is both open, so you can see what the “weights” that make it work are, and it’s flexible. The new Inkling model thus gives you a hint about how you and your company might soon start using next-generation AI, which could be much more crafted to meet your specific needs.

What should you be thinking about Inkling?

Inkling was released yesterday, and on its webpage Thinking Machines promises its “mission is to build AI that extends human will and judgment.” More specifically, it calls Inkling “a model we trained from scratch with the full weights available, so that people can make it their own.”

Here’s what “full weights available” actually means: For generative AI models like Inkling, ChatGPT, and many rivals, weights are essentially a vast array of numbers that constitute the training data set that makes the AI magic work. If you train a model with particular words over and over, the “weights” in the model linked to these words get tweaked, which influences the answers the AI gives to questions later on. For many mainstream AI models, these weights are proprietary and secret. When you query such models, for a vast variety of purposes—from writing code to drumming up a quick blurb for your website—you have to blindly trust that they’re working the way you want them to, and hope that the output is a good fit for your needs.

This is where Inkling stands out. Through a special tool called Tinker, users can shape and customize how Inkling works so that it aligns more precisely with their individual needs—which means you can tweak the model’s weights so that its outputs are a better fit to what you want.

Thinking Machines explains that Inkling is “designed to be broad,” and has been trained “across agentic, reasoning, coding, instruction-following, factuality, vision, and audio tasks, rather than narrowly optimizing for one domain.” This makes it ideal for “customization and real-world use: different users need models that can adapt to very different workflows,” the company says, arguing that this idea is better than an AI excelling on mere benchmark tests.

There’s also some protection built into the model, and the website explains that “the optimal policy is to answer when confident and otherwise say ‘I don’t know’ or give a hedged best guess.” That may boost your trust in this kind of tool, especially if you’ve read how AI can confidently hallucinate fake results, or you’ve seen how ChatGPT and its ilk have a problem when they lie about how many “rs” are in the word “strawberry.”

Inkling and next-gen AI

It’s worth pointing out that an inkling is a vague idea or a hint you have about something. Inkling (the model) is a first effort, and thus is probably a little rough around the edges. Right now you’ll also need to be familiar with the technical ins and outs of how AI works in order to shape Inkling to what you need.

You could also see Inkling as a hint about the way AI may go in the future. Right now AIs like Claude or even Siri are closed models, designed to be very general purpose. But in the future, companies like Thinking Machines may offer models that you can tailor in a more open way, even sharing your experiences in training them with other people. You can see how this would be useful: if you run an online book company, you don’t necessarily need to rely on (and pay for) a vastly complex, generally capable AI tool that also knows tips on how to machine metal objects on a lathe, or the history of French Impressionist painting. 

Odds are that these kinds of tailored models will align better with what your company needs. They could even be cheaper to run—some companies have landed themselves in the spotlight recently for the bank-breaking AI bills they’ve run up, or even the expensive way they incentivize workers for their AI use. Inkling gives us a hint about Thinking Machines intentions, with the website noting it “supports controllable thinking effort, allowing you to balance performance with token efficiency.” Tokens are the way AI firms measure units of use, so the plan here is that you don’t waste tokens by using inefficient models.

Take-aways for your company

Remember that this tool is new, and as yet you’ll need a certain level of skill to make the most of the way Inkling works. You’ll also have to pay to use it, much like you do for rivals like ChatGPT or even Apple’s new Siri AI. 

But if your team is technically capable, you may find Inkling and Tinker give you a sample of how future AIs will bend around your company’s needs, rather than requiring you to shape your business processes around the AI in certain ways, or even to budget for unexpected AI bills. 

Check out more at the Inkling site.

This post originally appeared at inc.com.

“Click here to subscribe to the Inc. newsletter: inc.com/newsletters"