Inkling: Thinking Machines Launches Open Multimodal AI

Thinking Machines Lab has launched its first public AI model, called Inkling. This new system features 975 billion parameters and understands text, video, and audio. It is one of the most ambitious open-source AI releases this year. What's interesting is Inkling is a multimodal model. It can process different types of information simultaneously. For example, you can give it a video clip, an audio recording, or a document, and it can reason across all these inputs. The 975 billion parameters make it very capable, especially compared to many models in the 7 to 70 billion parameter range. The company has been developing its AI infrastructure mostly in secret for about a year and a half. Now, Inkling is open source, which means developers can use and build on it without fees. Thinking Machines aims to create specialized AI for different industries, moving away from a "one-size-fits-all" approach. This strategy could give them a strong position against larger competitors. Inkling's ability to natively understand video and audio sets it apart from many other open models. This development could significantly impact how specialized AI tools are created and adapted across various sectors.