Mira Murati's Thinking Machines Releases Inkling — and Open-Weight AI Gets Its Moment
The buzziest startup of the post-OpenAI diaspora just shipped its first model, and developers can download and customize it without a proprietary API.
Thinking Machines, the startup founded by former OpenAI CTO Mira Murati, has released its first model — an open-weight system called Inkling that developers can download, fine-tune and deploy without relying on anyone's proprietary API.
The release lands at a moment when the industry is polarizing between closed frontier labs and an increasingly capable open ecosystem. By choosing open weights for its debut, Thinking Machines is planting itself firmly on the side of developer control.
Why open weights matter now
Enterprises have grown wary of building core products on APIs whose prices, policies and models change without notice. Open-weight models can run on private infrastructure, be audited for behavior, and be customized deeply for a domain — advantages that often outweigh a few points of benchmark performance. Even closed-lab executives have begun advocating for government oversight of frontier systems, implicitly acknowledging that raw capability is no longer the only axis of competition.
The competitive picture
Inkling joins a crowded open field alongside Meta's Llama family and strong Chinese entrants. What distinguishes it is pedigree: a founding team drawn from the people who built GPT-4-era systems. If Inkling's quality matches its lineage, the message to the market is uncomfortable for incumbents — the knowledge to build frontier AI has left the building, and it is publishing weights.
