np.random.seed(123)

N= 10
df = pd.DataFrame({
  'grp':np.random.choice(['aaa','bbb','ccc'], N, p=[0.3, .6, .1]),
  'val':np.random.randint(100, size=N)
})

In [119]: df
Out[119]:
   grp  val
0  bbb   46
1  aaa   96
2  aaa   25
3  bbb   83
4  bbb   78
5  bbb   36
6  ccc   96
7  bbb   80
8  bbb   68
9  bbb   49

In [120]: df['grp'].value_counts()
Out[120]:
bbb    7
aaa    2
ccc    1
Name: grp, dtype: int64

idx = (df.groupby('grp')['val'].transform('count')).sort_values(ascending=False).index

df.reindex(idx)

In [126]: df.reindex(idx)
Out[126]:
   grp  val
9  bbb   49
8  bbb   68
7  bbb   80
5  bbb   36
4  bbb   78
3  bbb   83
0  bbb   46
2  aaa   25
1  aaa   96
6  ccc   96