Skip to content

Individual Runs (2026-07-06)

Seven models/configurations tested using llama-benchy v0.4.0 on the same stack.

ModelConfigKey result
Ornith-IQ4 (IQ4_XS-MTP-graft)IQ4_XS + Q6_K head−18% smaller, +10% faster, better KLD — new prod candidate
Ornith-1.0-35B (prod)Q4_K_M + MTP n2Flat ~76 t/s 0→16k, shallow −24% decay to 65k
Qwen3.6-35B-A3B vocab-trimmedQ4_K_S + MTP n2Shallowest decay (−19%), pulls ahead at deep context
Qwen3.6-35B-A3B MTP OFF vs ONQ4_K_MCleanest isolation: MTP = +58% at 65k
Qwen3.6-35B-A3B DFlash vs MTPQ4_K_MMTP wins (92% accept vs ~60%), DFlash stable with -ub 512
gemma-4-12b-it (dense)Q4_K_M + MTP n3Dense penalty: ~2× slower than A3B models
Qwen3.6-27B (dense)Q4_K_M + MTP n2Slowest model (23 t/s); MTP net-neutral on dense