InclusionAI's fast hybrid-reasoning MoE, 560B total with 25B active.
Add a second model to compare.
This model hasn't been benchmarked yet.