Testing the Depths of AI Empathy: Q4 2025 Benchmarks

Q4 2025 LLM empathy benchmarks are out, with Gemini 3 Pro, Claude Sonnet 4.5, and Kimi-2 Instruct scoring perfectly or near-perfectly, suggesting they may be "gaming the test" by knowing the correct answers for the EQ-60 and SQ-R. Kimi-2 Instruct is noted for its exceptional speed and high quantitative and qualitative scores. Future focus will shift from quantitative scores to assessing perceived empathy in multi-turn chat conversations, as high scores don't always correlate with natural, empathetic dialogue.