Modified version which compares the single method check times (which need to include Mock.instance_methods(false)).

respond_to_vs_include.rb

# Fancy benchmark...
# is respond_to? so faster than include?

require 'benchmark'

n = 10_000_000
mock_methods = ('a'..'z').to_a

class Mock
end

mock_methods.each do |m|
  Mock.class_eval do
    send(:attr_accessor, m)
  end
end

mock = Mock.new
instance_meths = Mock.instance_methods(false)

Benchmark.bm(15) do |test|
  test.report('respond_to?: ') do
    n.times do
      mock.respond_to? :k
    end
  end
  test.report('include?: ') do
    n.times do
      instance_meths.include? :k
    end
  end
  test.report('instance_methods.include?: ') do
    n.times do
      Mock.instance_methods(false).include? :k
    end
  end
end

times.txt

$ ruby respond_to_vs_include-RK.rb
                                               user     system      total        real
respond_to?:                               1.170000   0.000000   1.170000 (  1.167067)
include?:                                  7.862000   0.000000   7.862000 (  7.880450)
instance_methods.include?:               152.242000   0.000000 152.242000 (152.295711)