optimizeModel

Whether to allow model optimization techniques such as quantization, speculative decoding, and kernel tuning. The default is true.