Thinking Machines Lab releases Inkling, 975B-parameter open model
Mira Murati's lab dropped Inkling: a Mixture-of-Experts transformer with 975B total parameters (41B active), multimodal, trained on 45 trillion tokens of text, images, audio, and video.
Apache 2.0 licensed and immediately available. A smaller variant (276B total, 12B active) called Inkling-Small is in testing; weights come later.
Model card and training data docs are sparse compared to typical US lab releases — minimal transparency on dataset sourcing.