Last updated 2 September 2009.

At this time we have five sets of benchmark data online:

• Our genuine R6RS benchmarks consist of portable R6RS top-level programs.
• Our fake R6RS benchmarks are written in R5RS Scheme, and test the performance of a wide variety of popular implementations.
• We have a very few benchmarks for an early version of Common Larceny.
• Our older benchmark results may still be of some interest because they include some timings for C, C++, Java, and SML/NJ.
• The Kernighan and Van Wyk micro-benchmarks were written to test the performance of a variety of scripting languages, including Scheme. They include several good tests of file i/o, and have been incorporated into our fake R6RS benchmarks and have been improved upon in our genuine R6RS benchmarks.