GPT-5.5 vs Claude vs Gemini: Which AI Explains Things Clearest? I Built a Benchmark to Find Out

This is a report on a Kaggle Benchmarking Challenge that aimed to determine which AI model explains complex topics most clearly. The author built a benchmark with eight test prompts covering subjects like recursion, photosynthesis, cryptocurrency, and machine learning. The goal was to identify which AI assistant would provide the clearest answer for beginners, as the wrong model could result in a confusing response.

Three AI models were tested: GPT-5.5 from OpenAI, Claude Sonnet 4.6 from Anthropic, and Gemini 3.7 Flash from Google. The findings showed that GPT-5.5 scored 76.3%, significantly outperforming both Claude and Gemini, which both scored 38.8%. The author was surprised by the results, as GPT-5.5 almost doubled the scores of the other two models.

The scoring system rewarded responses that were between 50-200 words long - a balance between being concise enough to be clear and detailed enough to be useful. However, GPT-5.5 consistently achieved this balance, while Claude and Gemini tended to over-explain, leading to lengthy responses that could overwhelm a beginner. The conclusion drawn was that more words does not necessarily equate to a better explanation.

The author noted that while the benchmark provides a data-driven comparison, it might be beneficial to involve real beginners in the evaluation process to measure clarity directly, as human judgment could yield different results. The full benchmark can be accessed at https://www.kaggle.com/benchmarks/michaelomijiemkings/who-explains-things-clearest.