Inkling AI Model Released with Open Weights for Customization

Inkling is a Mixture-of-Experts transformer model with 975 billion total parameters and 41 billion active parameters, capable of handling a context window of up to 1 million tokens.

The model was pretrained on 45 trillion tokens of diverse data, including text, images, audio, and video. Inkling-Small, a lighter version with 12 billion active parameters, is also being released for lower-cost and low-latency applications.

Inkling supports multimodal reasoning and is designed to be a flexible foundation model, suitable for various domains and workflows. It is available for fine-tuning on the Tinker platform, which includes a new Inkling Playground for developers to interact with the model.

The model has demonstrated its capabilities by successfully creating a functional web application and a polished travel journal, showcasing its potential for real-world applications.

Inkling allows developers to control thinking effort, balancing performance and efficiency, which is crucial for applications requiring low latency and cost-effectiveness. The model is part of a broader initiative to enhance collaborative AI interactions.