# 32B inference optimization

A 32-billion-parameter dense language model focused on enhancing reasoning capabilities, built upon Qwen 2.5-32B-Base, demonstrating performance comparable to larger MoE models in reasoning benchmarks.