I ran the 35B agentic comparison someone asked for (stock vs Ornith vs KAT-Coder, 120 runs)

Someone in the comments of my 27B post-train bakeoff asked for the 35B version, so I ran it. Same setup as last time: fresh Coder workspaces on my k8s cluster, each driving my own agent (Hermes) headlessly, models on llama.cpp via llama-swap on one 5090, every call traced through an OTel shim into S

中文全文翻译正在补齐，可先查看下方英文原文。

Someone in the comments of my 27B post-train bakeoff asked for the 35B version, so I ran it. Same setup as last time: fresh Coder workspaces on my k8s cluster, each driving my own agent (Hermes) headlessly, models on llama.cpp via llama-swap on one 5090, every call traced through an OTel shim into S