Anthropic’s Claude 3.7 Sonnet was the best performer, managing to successfully debug the faulty code in 48.4% of cases. OpenAI’s o1 achieved success 30.2% of the time, while OpenAI’s o3-mini ...

Bard can now generate code, debug existing code, and help explain lines of code. Bard can now generate code, debug existing code, and help explain lines of code. Tom Warren is a senior editor and ...