OpenAI: o4 Mini High

OpenAI's o4 Mini High, released on April 16, 2025, is a compact reasoning model designed for efficiency and strong multimodal capabilities, with its reasoning effort set to high. It exhibits moderate speed performance, ranking in the 31st percentile, and is positioned at premium pricing levels, in the 14th percentile. A standout feature is its exceptional reliability, achieving a 100% success rate across all benchmarks. The model demonstrates perfect accuracy in General Knowledge, making it the most accurate model at its price point and among models of similar speed. It also excels in Reasoning (98% accuracy, 92nd percentile) and Coding (93% accuracy, 84th percentile), outperforming many peers. Its Email Classification accuracy is strong at 99% (76th percentile). While its Ethics performance is respectable at 98% (38th percentile), its Instruction Following accuracy is 71% (82nd percentile). A notable weakness is its performance in the Hallucinations benchmark, where it achieved only 62% accuracy (12th percentile), indicating a tendency to not appropriately acknowledge uncertainty. Overall, o4 Mini High is a highly reliable model with strong reasoning and coding capabilities, particularly suited for high-throughput scenarios where accuracy in complex tasks is critical, despite its premium cost and moderate speed.