“In NVIDIA's ENPIRE experiments, models based on GPT-5.5 and Opus 4.7 demonstrated the best performance, while Kimi-2.6 lagged behind.”