Thinking Machines Lab releases Inkling, a 975B-parameter open-weight multimodal mixture-of-experts model

Thinking Machines Lab/Inkling

Thinking Machines Lab released Inkling, an open-weight multimodal language model with a mixture-of-experts architecture. The model has 975B total parameters with 41B active per token, demonstrating that frontier-quality models can be released openly to the research and builder community.

• 975B total parameters, 41B active per token through sparse expert routing.
• Multimodal: processes text, images and other data in a unified architecture.
• Published as open-weight under a permissive license.
• Smaller variant (Inkling-Small) with 276B total parameters also available.