Untouched Qwen3.8-27B served from our own vLLM endpoint — the control for the post-training experiments. Answers here are the baseline every adapter is compared against.
Notes
Self-hosted on our vLLM endpoint as "qwen3.8-27b" (TUNED_LLM_BASE_URL). LoRA experiment model — not on OpenRouter. Warm the endpoint before a session; first request after idle pays the cold start.