In datasets where the target variable spans more than one order of magnitude like yours, the default
RMS error metric puts too much emphasis on the large points, which can make the search slow.
In cases like this, the
Hybrid (CC+RMS) metric can be faster. Using this metric, I found the following solution for your dataset in only a few minutes:
y = -10.9336+exp(tan(sin(tan(tan(0.00924664*x))-0.0891209))+3.48883)