A smaller, efficient variant of GPT-5.4 optimized for low-latency tasks with reduced computational cost.