Kimi K3 dominates frontend code, but fails at complex math

moonshot's kimi k3 has become the first chinese model to claim the top spot on the code arena: frontend benchmark, scoring 1,679 based on human preferences and outperforming rivals like claude fable 5 and gpt-5.6 sol. despite this coding prowess, the model struggles with complex mathematical reasoning, achieving only 39% accuracy on frontiermath tier 4's hardest tasks compared to nearly 90% for models from openai and anthropic. this divergence highlights a specialized performance profile, where kimi k3 excels in practical coding applications but lags in abstract problem-solving domains.