Inkling 2 Mini releases open weights, using a recurrent architecture for local inference
Inkling 2 Mini
Inkling 2 Mini is out with open weights: a 26.4B-parameter MoE built on a recurrent Transformer, at roughly 20GB of total memory. Each token runs the middle layers three times. The publisher reports 48.2% on DeepSWE v1.1 and 46.5% on Vals Index v2.2, and says the model trails Qwen3.8-27B on multi-turn dialogue and multi-constraint instructions.
Why it matters: Inkling 2 Mini runs its middle layers repeatedly for efficiency, and its experiments separate how recurrence affects multi-step reasoning versus factual knowledge.