---
title: Tinkersnot Bench Server
emoji: 🦙
colorFrom: blue
colorTo: green
sdk: docker
app_port: 7860
pinned: false
---

OpenAI-compatible llama.cpp server for base-model bake-off eval.
Endpoint: `/v1/chat/completions`. Model set via `LLAMA_ARG_HF_REPO` in the Dockerfile.