Reddit discussion seeks MoE models with ~2B active parameters

A Reddit user asks for Mixture-of-Experts models with around 2B active parameters, noting a gap between existing 1B-active models (LFM2.5 8B A1B, Granite 4.0h 7B A1B) and 3B+ active models (Qwen 3.x ~30B A3B, Gemma 4 26B A4B). The thread has 32 upvotes and 23 comments.