Cheers. It'd be useful if you indicated what the numbers meant (seconds, other metrics) and whether lower or higher is better. Also, when you say Opera Mini, do you really mean Opera Mini or do you mean Opera Mobile? Both run on the N900 but Opera Mobile would be the most relevant to test as it's the only one that has a Maemo version that can be installed by end-users without having to rely on hacks. EDIT: just saw your reply above - so it's Opera Mobile 10 you've tested then.
Thanks for the good work, I'm quite surprised as well, and even with supposed PR(L)1.2 speed bump, micro-B came in last. A few questions though, How did you calculate the load time of the pages? If you'd like to make a compete benchmark I think that loading these pages with the cache persisting may how some interesting results as well for oft visited pages.