This NVIDIA-backed AI lab just dropped its first model — why Anthropic should be worried

Thinking Machines Lab quietly dropped its new open-weight Inkling AI model in a move that could make the AI race at the frontier that much more interesting. Indeed, before the big Inkling drop, it was easy to forget about Mira Murati's AI startup.

The former OpenAI CTO is certainly going about AI innovation from a vastly different angle. And I do think that the leading AI labs, especially Anthropic, which took the world by storm with Claude Code, Mythos, and, now, Fable, are going to need to stay on their toes to keep up the pace as the number of fierce, scary competitors with their unique moats look to play to win.

In any case, there's been no shortage of criticisms surrounding closed-source models as well as the power possessed by the AI labs at the frontier. Add the uncertainty about what's "too dangerous" to release and guardrails placed in Claude Fable 5 into the equation, and it certainly seems like Thinking Machines Lab has an opportunity to capture some of the enterprise users who might be tempted to jump ship.

Inkling debuts — it's a serious competitor at the frontier

Like Anthropic's models, Thinking Machines Lab is serious about efficiencies and other innovations over sheer scaling. As the AI lab looks to shift the business model, it might also change the landscape in a profound way as firms grow tired of spending on the so-called "token tax." In any case, time will tell how Inkling fairs and, more importantly, what it evolves to become as the AI race moves into an interesting new phase, one that could prioritize efficiencies as much as raw firepower.

As the battle between the cloud, edge, and on-prem unfolds, and closed-source dukes it out with open source and now open-weights, it's going to be very interesting to see how the market reacts. In any case, Thinking Machines Lab definitely stands out as an AI lab that could further diversify one's portfolio whenever the firm decides to go public.

In the meantime, Nvidia (NASDAQ:NVDA) and Thinking Machines Lab have a long-term deal, inked earlier in the year, in place as the open-weight model maker looks to "advance the frontier of AI." The 1-GW strategic partnership might just cement Muri Murata's startup as a credible threat in the AI race, as it gets its hands on enough compute to move ahead with the astronomical amount required to train next-generation models.

Thinking Machines Labs is innovating in its own way. It's worth keeping tabs on.

With such a big deal in place, it certainly feels like Thinking Machines Lab has an infrastructure moat that many other rivals hoping to move to the front of the AI race might lack. At this juncture, it's hard to tell if open-weights or closed-source is the future.

Sure, customers might have gripes about closed-source, but either way, Nvidia seems to have many of its bases covered, regardless of what the future holds for AI and the architecture that firms will be willing to take on.

With Anthropic picking up serious traction, I do think closed-source could continue to dominate, but if distaste for token tax mounts, we could eventually see a shift. In any case, whether Nvidia's deal and stake in Thinking Machines Lab is a "kingmaker" move remains the trillion-dollar question. I'd chalk it up as another brilliant bet by Jensen Huang and his team.

In any case, Anthropic might wish to be a bit more mindful about how they stretch their pricing power. Their financials are in much better shape than the likes of OpenAI, but, at the same time, new entrants with their own unique advantages might take share if enterprises believe the token tax has gotten out of hand.

The bottom line

For now, Mythos feels untouchable, and if the unprecedented capabilities continue to deliver immense value, the firm can charge what it wants. I guess it all comes down to what a firm prioritizes: strategic smarts to gain a competitive advantage over rivals that only the very best model can provide, or "good enough" models that can save significant sums in the long run.

My guess is the former wins out in these earlier innings of AI. It's profound, even unfathomable, to even begin to comprehend what Mythos has accomplished already regarding spotting vulnerabilities in software. For that kind of value provided, I'd say Anthropic deserves to and should charge top dollar.

Contact [email protected] for any questions or corrections.