Anthropic: Claude Opus 4.7

Anthropic's Claude Opus 4.7, created on April 16, 2026, is designed for long-running, asynchronous agents, building on its predecessors' coding and agentic strengths. This model performs among the fastest models, ranking in the 61st percentile for speed across benchmarks. However, it is positioned at premium pricing levels, ranking in the 8th percentile for cost. A significant strength is its exceptional reliability, boasting a 99% success rate. Opus 4.7 demonstrates strong performance across various benchmarks. It achieves perfect accuracy in General Knowledge and Ethics, notably being the most accurate model at its price point and speed in these categories. Its Coding and Instruction Following capabilities are also impressive, ranking in the 95th and 89th percentiles respectively. Reasoning and Mathematics also show high accuracy at 98% and 94%. While its Hallucinations accuracy is 90%, placing it in the 37th percentile, this indicates a tendency to provide answers rather than acknowledge uncertainty in fictional scenarios. Email Classification is solid at 98% accuracy. Overall, Opus 4.7 excels in complex cognitive tasks and coding, making it a powerful tool despite its premium cost.