Mira Murati left OpenAI. Then she raised $2 billion. Now she’s giving away the goods.
Thinking Machines Lab, the AI startup founded by OpenAI’s former CTO, launched its first open-weight model on July 15. It’s called Inkling, and it’s not small. The model packs 975 billion total parameters in a Mixture-of-Experts architecture, with 41 billion active at any given time. It processes text, audio, and video, and supports a context window of up to 1 million tokens.
Murati is releasing it under an Apache 2.0 license, meaning developers can download, modify, and deploy it with virtually no strings attached.
What Inkling actually is
The Mixture-of-Experts approach is worth unpacking. In English: instead of firing up all 975 billion parameters for every query, the model activates only a relevant subset, roughly 41 billion parameters, depending on the task. It keeps compute costs manageable while preserving the depth of a much larger model.
