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

Thinking Machines Lab released Inkling, an open-weights artificial intelligence model with 975 billion total parameters that can process text, images, and audio as inputs while producing text outputs. The model uses a Mixture-of-Experts architecture where only 41 billion parameters are active at any given time, allowing it to balance capability with efficiency. A key differentiating feature is "controllable thinking effort," which lets users adjust how many tokens the model spends reasoning about each task, making both cost and latency tunable rather than fixed. The model was trained on 45 trillion tokens of multimodal data and performs competitively on various benchmarks, with deployment options available through multiple frameworks and hosted APIs.