OpenAI: GPT-5

OpenAI's GPT-5, released in August 2025, demonstrates a strong focus on advanced reasoning and accuracy for complex tasks. While its speed performance is moderate, ranking in the 32nd percentile, it consistently delivers exceptional reliability with a 100% success rate across all benchmarks, indicating minimal technical failures. GPT-5 is positioned at a premium price point, falling into the 10th percentile for cost. The model excels in several critical areas. It achieves perfect accuracy in both General Knowledge and Ethics, with the latter also being the most accurate among models of comparable speed and price. Its Instruction Following capabilities are outstanding, ranking in the 96th percentile and noted as the most accurate at its price point. Reasoning and Coding also show very strong performance, with 91st and 83rd percentile accuracy respectively. Hallucination rates are commendably low at 98% accuracy, indicating a strong ability to acknowledge uncertainty. Email Classification and Mathematics show solid, though not top-tier, performance. GPT-5's strengths lie in its robust reasoning, precise instruction following, and high accuracy in knowledge-based and ethical tasks, making it well-suited for high-stakes applications requiring meticulous output.