MERLIN-2.1 is an adaptation of the MERLIN-2.0 package, mainly designed to run on computers using 32-bit floating point arithmetic. The standard Merlin-2.0 on such machines, achieves a precision of 7 significant digits at most. This is insufficient since in many real problems a higher precision is required. MERLIN-2.1 treats this inadequacy using double precision operations, enhancing so the precision up to about 14 significant digits, in the standard user-friendly Merlin environment. © 1990.