Benchmarks are basically tied https://artificialanalysis.ai/models/comparisons/deepseek-v4...