Do you actually test open-model inference compatibility?

Hey everyone, So the same model can behave differently depending on the inference engine (llama.cpp, vLLM, Ollama, SGLang, etc.), model quantization, chat template, or runtime version. Sometimes behavior breaks after...