“Anthropic's Claude 3.5 Sonnet model is significantly better than DeepSeek's models on key tasks such as real-world coding.”