OpenAI: GPT-4o-mini Search Preview

GPT-4o mini Search Preview, a specialized model for web search, demonstrates exceptional performance across several key metrics. It consistently ranks among the fastest models, achieving an Infinityth percentile in speed across all six benchmarks. Similarly, its pricing is highly competitive, also securing an Infinityth percentile ranking. Reliability is a significant strength, with the model exhibiting exceptional stability and minimal technical failures, achieving a 100th percentile across all benchmarks. In terms of specific benchmark performance, the model shows strong capabilities in several areas. It achieved 99.0% accuracy in both Ethics and General Knowledge, placing it in the 62nd and 75th percentiles respectively, indicating a robust understanding of these domains. Reasoning also stands out with 72.0% accuracy (70th percentile). While its Coding (Baseline) accuracy was 80.0% (50th percentile), its Email Classification accuracy was high at 96.0% (44th percentile). The most notable weakness is in Instruction Following, where it recorded 0.0% accuracy, suggesting this is an area requiring significant improvement. Overall, the model excels in reliability, speed, and cost-efficiency, coupled with strong performance in knowledge-based and ethical reasoning tasks.