Narval is a proprietary large language model (LLM) with 4 billion parameters, pre-trained on financial discussions from X (Twitter) to perform advanced Name-Entity Recognition and Relation Extraction tasks. The model is capable of precisely analyzing economic and financial reasoning, offering a nuanced understanding of the mechanisms and concepts that shape this domain, while accurately capturing the tone of discourse — from pessimism and negativity to neutrality and optimism — with a level of consistency and precision that surpasses that of general-purpose generative LLMs.