Thinking Machines Lab Releases Inkling: A 975B-Parameter Open-Weights Multimodal MoE With 41B Active Parameters And Controllable Thinking Effort

Short summary

Thinking Machines Lab released Inkling, a 975B-parameter open-weights Mixture-of-Experts multimodal model with 41B active parameters and a 1M-token context window, shipped under Apache 2.0. The lab explicitly positions it as a customization base rather than a frontier leader, with controllable thinking effort as its key differentiator. It supports native text, image, and audio input.

• •975B-parameter MoE with 41B active params, 1M-token context, Apache 2.0 license
• •Positioned as a customization base, not a frontier model
• •Controllable thinking effort is the practical differentiator

Generated with AI, which can make mistakes.