Patronus GLIDER is a fine-tuned phi-3.5-mini-instruct model that serves as a general evaluation tool, judging text, dialogue, and RAG settings according to user-defined standards and scoring rules. Trained on synthetic and domain-adaptive data, it encompasses 183 metrics and 685 domains, including finance and medicine. The model supports a maximum sequence length of 8192 tokens, but tests have shown it can handle longer texts (up to 12000 tokens).