Here's an even simpler benchmark that shows the difference between throw and throwIO: {-# LANGUAGE ScopedTypeVariables #-} import Control.Exception import Prelude hiding (catch) import Criterion.Main ignoreExceptions :: IO () -> IO () ignoreExceptions m = m `catch` \(_ :: SomeException) -> return () main :: IO () main = defaultMain [ bench "throw" $ ignoreExceptions (throw DivideByZero) , bench "throwIO" $ ignoreExceptions (throwIO DivideByZero) ] $ ghc --make Throwing.hs -O2 -fforce-recomp -o throwing [1 of 1] Compiling Main ( Throwing.hs, Throwing.o ) Linking throwing ... $ ./throwing warming up estimating clock resolution... mean is 21.65454 us (40001 iterations) found 1884 outliers among 39999 samples (4.7%) 412 (1.0%) high mild 1462 (3.7%) high severe estimating cost of a clock call... mean is 1.893442 us (42 iterations) found 4 outliers among 42 samples (9.5%) 1 (2.4%) high mild 3 (7.1%) high severe benchmarking throw collecting 100 samples, 258616 iterations each, in estimated 2.165460 s bootstrapping with 100000 resamples mean: 85.76940 ns, lb 85.20807 ns, ub 86.40616 ns, ci 0.950 std dev: 3.047974 ns, lb 2.681299 ns, ub 3.679702 ns, ci 0.950 found 1 outliers among 100 samples (1.0%) variance introduced by outliers: 0.999% variance is unaffected by outliers benchmarking throwIO collecting 100 samples, 319584 iterations each, in estimated 2.165460 s bootstrapping with 100000 resamples mean: 62.71774 ns, lb 62.41275 ns, ub 63.45298 ns, ci 0.950 std dev: 2.274746 ns, lb 1.130328 ns, ub 4.596054 ns, ci 0.950 found 9 outliers among 100 samples (9.0%) 4 (4.0%) high mild 5 (5.0%) high severe variance introduced by outliers: 0.999% variance is unaffected by outliers Regards, Bas