Unsloth, Swift1.5, Peculiar-Ragdoll, ThinkingCap - Qwen3.8-27B

A comparison of Qwen3.8-27B checkpoints, including Swift1.5, Peculiar-Ragdoll, Unsloth Q4_K_XL, and ThinkingCap Q4_K_M, was conducted using a 69-question evaluation set. An additional test was performed on mradermacher/Signal-3.8-27B-Terse-Coder-i1-GGUF (Q4_K_M quant) using llama.cpp due to tensor incompatibility with a Splash-based engine, noting that the total time taken for this specific test was not a fair comparison.