Thinking Machines released Inkling 2 Mini, a 26B parameter AI model under an Apache 2.0 license that is small enough to run on Mac laptops with 32GB of RAM. The model uses a looped transformer architecture and scored 20.1 on the AA Index, with some benchmarks showing it ties the performance of Claude Opus 5.5 on CyberBench.
The architecture consists of 24 layers where the middle 12 loop three times to allow the model to reason more while generating less text. To support study into the monitorability of these internal mechanisms, Thinking Machines also released a variable-loop research model. The company stated that Inkling 2 Mini outperforms its July 2026 predecessor, Inkling 1.0, while being only 2.7% of its size.
