If you read the golang-dev google groups mailing list you will see lots of micro-benchmarks posted for specific commits. The best test is to install both stable and tip on your machine and compile your application with both. In one of my application there was a 9% improvement, in another it was 53%. When Go 1.1 is released I presume the computer benchmarks game will be updated, which might make comparison, at lease with other languages, easier.