Thinking Machines Lab, the AI startup founded by former OpenAI CTO Mira Murati, released its first in-house AI model Wednesday. Called Inkling, it is an open-weight mixture-of-experts system with 975 billion total parameters, though it only activates about 41 billion per task.
Trained on 45 trillion tokens spanning text, image, audio, and video, Inkling reasons natively across all four modalities but currently outputs only text, code, and structured data. The model lets users dial “thinking effort” up or down depending on speed requirements and is designed to flag uncertainty rather than guess.
On one benchmark, Thinking Machines says Inkling uses a third as many tokens as Nvidia’s Nemotron 3 Ultra to hit the same coding performance. The model is positioned as a test of the company’s central bet: that AI organizations can adapt themselves will outperform the one-size-fits-all models sold by the biggest labs.
This is the company’s first public proof point after a year and a half of building AI infrastructure largely out of public view. Inkling joins a growing roster of open-weight alternatives challenging the dominance of proprietary frontier models.
