diff --git a/Number_types/doc/Number_types/CGAL/Gmpzf.h b/Number_types/doc/Number_types/CGAL/Gmpzf.h
index 876bd37069f..675e55ac9e4 100644
--- a/Number_types/doc/Number_types/CGAL/Gmpzf.h
+++ b/Number_types/doc/Number_types/CGAL/Gmpzf.h
@@ -9,7 +9,8 @@ numbers of the form \f$ m*2^e\f$, where \f$ m\f$ is an arbitrary precision integ
 based on the \gmp library, and \f$ e\f$
 is of type `long`. This type can be considered exact, even if the
 exponent is not a multiple-precision number. This number type offers
-functionality very similar to `MP_Float` but is generally faster.
+functionality very similar to `MP_Float` but is generally faster. `Mpzf` also
+provides similar functionality and is generally faster than `Gmpzf`.
 
 \cgalModels{EuclideanRing,RealEmbeddable}
 
diff --git a/Number_types/doc/Number_types/CGAL/MP_Float.h b/Number_types/doc/Number_types/CGAL/MP_Float.h
index 6f3b097d71b..ac2c6124b75 100644
--- a/Number_types/doc/Number_types/CGAL/MP_Float.h
+++ b/Number_types/doc/Number_types/CGAL/MP_Float.h
@@ -24,8 +24,8 @@ plan to also have a multiprecision exponent to fix this issue.
 
 The implementation of `MP_Float` is simple but provides a quadratic
 complexity for multiplications. This can be a problem for large operands.
-For faster implementations of the same functionality with large integral
-values, you may want to consider using `GMP` or `LEDA` instead.
+For faster implementations of the same functionality, if `GMP` is available,
+you may want to consider using `Mpzf` or `Gmpzf`.
 
 */