Qwen3.8 27B Huihui-Abliterated
on NVIDIA RTX 5090 · NInfer · 184,320 ctx
Sep 7, 2026
Use cases
agenticcoding
Summary
User benchmarks 180K context with MTP5 enabled, reaching 2587 t/s prefill and 200.5 t/s decode.
The run also covers 64K and 120K contexts.
A synthetic corpus yields high MTP acceptance, while real agentic coding logs show about 51% accept and about 154 t/s.