Anthropic’s Opus 5 achieved a perfect score of 42/42 on the 2026 International Mathematical Olympiad problems, according to AI Grid. The model outperformed its predecessor Claude Fable 5 in key benchmarks like ARC AGI 3, scoring three times higher than its closest competitor on that benchmark. Opus 5 offers superior or comparable performance to Claude Fable 5 at half the cost. It excels in practical applications such as coding, financial modeling, legal document review and medical diagnostics. However, the model introduces trade-offs, including diminished performance in higher-level reasoning tasks and a fallback mechanism that reverts to the Opus 4.8 model under certain conditions, which can lead to performance inconsistencies in critical tasks. Minor misaligned behaviors and limitations in high-level reasoning and complex tasks highlight areas for cautious use.