Text this: Comparative Evaluation of GPT-4o, GPT-OSS-120B and Llama-3.1-8B-Instruct Language Models in a Reproducible CV-to-JSON Extraction Pipeline.