Inkling 2 Mini
Hugging Face released Inkling 2 Mini, a compact AI model with 26.4B parameters designed to run on laptops while delivering high performance on coding tasks. Using a looped transformer architecture that processes layers multiple times per token, Inkling 2 Mini outperforms all open-weight models under 100B parameters on benchmarks like DeepSWE v1.1 (48.2%) and runs faster than comparably performing models. The looping mechanism allows the model to reason longer while maintaining interpretability through shorter but legible chain-of-thought outputs.