The recent release of LuaJIT 2 establishes LuaJIT as the fastest dynamic language implementation. 32/64 bit Intel and ARM processors are supported, and performance can be comparable to compiled code.
There are some differences you should be aware of. LuaJIT is stricter than vanilla Lua and does not support the old Lua 5.0 way of dealing with a variable number of arguments.
LuaJIT now supports the dumping and reading of bytecode, but currently libraries such as Lanes cannot work with it currently. This applies to other extensions like Pluto which need deep integration with the Lua implementation.