Anthropic: Claude Sonnet 5

Anthropic's Claude Sonnet 5, released June 30, 2026, demonstrates strong frontier performance, particularly in specialized domains. Its speed ranking places it in the 46th percentile, indicating competitive response times, while its pricing is positioned at premium levels, ranking in the 16th percentile. A standout feature is its exceptional reliability, achieving a 100% success rate across all benchmarks, signifying minimal technical failures. The model exhibits notable strengths in Instruction Following (86th percentile), Coding (84th percentile), and Reasoning (87th percentile), aligning with its description as capable across coding, agents, and professional work. Its performance in Mathematics (76th percentile) and Email Classification (74th percentile) is also robust. While its General Knowledge (48th percentile) and Ethics (44th percentile) scores are solid, they are closer to the median compared to its top-tier performance in other areas. A key strength is its high accuracy in avoiding hallucinations (98.0%), demonstrating appropriate uncertainty acknowledgment. The model's cost per benchmark varies, with some tasks like Ethics and Email Classification being relatively cost-efficient despite its overall premium pricing.