gpt-5 competition

Anthropic’s Claude Opus 4.1 leads in coding performance with 74.5% accuracy on the SWE-bench test, surpassing OpenAI and Google, but…