Mira Murati spent years helping build some of the most powerful closed AI systems on the planet at OpenAI. Now she’s giving it all away for free.
Thinking Machines Lab, the AI startup Murati founded after departing OpenAI, released its first in-house model called Inkling on July 15. The model packs 975 billion parameters, processes text, images, audio, and video simultaneously, and ships under the permissive Apache 2.0 license. Anyone can download the weights from Hugging Face and do essentially whatever they want with them.
What Inkling actually is
Inkling uses a Mixture-of-Experts architecture, a design pattern where only a subset of the model’s total parameters activate for any given task. While the full model weighs in at 975 billion parameters, only 41 billion are active at any one time. The training run consumed 45 trillion tokens and was built from scratch on NVIDIA GB300 NVL72 hardware. The model supports a context window of 1 million tokens.
Thinking Machines Lab is also releasing a platform called Tinker alongside Inkling, designed to let researchers and developers customize the model for specific use cases.
