This study investigates how quantization affects reinforcement learning in large reasoning models, finding that post-training quantization and QLoRA outperform quantization-aware training.