Thinking Machines Releases Inkling 2 Small: A 594B MoE Model with 17B Active Parameters

On November 5, 2026, Thinking Machines released Inkling 2 Small. According to reports, the model utilizes a Mixture of Experts (MoE) architecture boasting 594 billion total parameters with 17 billion active parameters. Reported facts highlight its scale within the recent wave of MoE architectures. Analysis suggests that such designs aim to maximize model capacity while restricting active compute costs, though no hands-on testing was performed.