Snowflake managed Iceberg Table, the FLOAT type is written as 32-bit single precision in the file according to the iceberg FLOAT type specification. 32-bit floating-point number (float): Because fewer bits are used, the range of numbers that can be stored is narrower and the precision is limited to approximately seven decimal digits. Snowflake Managed Table In the case of Snowflake Table, the stored FLOAT type is always treated as 64-bit double precision. 64-bit floating-point number (double): Uses more bits (especially the exponent and mantissa parts) to represent numbers, so a very wide range of numbers can be stored with a decimal precision of about 15 to 16 digits.