The New American AI Model Designed to be Customized
From ByteByteGo · subscribed via aiste.ulozaite@gmail.com · original ↗ · unsubscribe
- Thinking Machines released Inkling, a 975B parameter model with only 41B active per token using Mixture of Experts architecture
- Model features 66 layers with 256 experts per layer (6 activated per token), mixed local/global attention, and supports 1M token context
- Includes customization tools for fine-tuning with user data, multimodal input support, and adjustable reasoning effort setting
Couldn’t fetch the full article — read it on the original site ↗.