Locations exceeding 1,000 millisieverts/hour, or 1 Sievert/hour, are marked in red circles. (They do not include measurements inside Containment Vessels.)
From TEPCO, data compiled on March 22, 2013 from the data taken between April 2011 and February 2013 (from TEPCO's webpage for plant survey maps):
Reactor 1, 1st floor: 4,700 millisieverts/hour (or 4.7 Sieverts/hour) where steam was gushing from the floor, locations over 1,000 millisieverts/hour nearby.
Reactor 1, 2nd floor: over 1,000 millisieverts/hour (or 1 Sievert/hour). Precise measurement was not done. The survey meter went overscale.
Reactor 2, 1st floor: (surface radiation level) 3,000 to 4,400 millisieverts/hour (or 3 to 4.4 Sieverts/hour) on the upper penetration hole on the Containment Vessel.
Reactor 2, 2nd floor: (probably surface) 330 to 1,640 millisieverts/hour (or 0.33 to 1.64 Sievert/hour) on X29B/C penetration hole.
Reactor 2 Turbine Building Basement: over 1,000 millisieverts/hour (or 1 sievert/hour).
Reactor 3, 1st floor: 2,290 millisieverts/hour (or 2.29 Sieverts/hour) along the rails for the equipment hatch. 4,780 millisieverts/hour (or 4.78 Sieverts/hour) on the floor surface.
From plant survey map as of 6/17/2013:
SGTS (Standby Gas Treatment System) pipe for Reactors 1 and 2: (surface) over 10,000 millisieverts/hour (or 10 Sieverts/hour)
Radiation measured on the 5th (top) floor of Reactor 3 (2,170 millisieverts/hour max) is in line with the levels measured on the 1st floor along the equipment hatch rails. As a reminder, that hatch was found open, and high radiation along the rails was from the steam or water leaking out from inside the Containment Vessel. Packbot did the cleanup of the hatch rails twice, and the towels that the robot used were soaking wet in the video from November 2011.
Packbot went in to clean the rails in November 2011, and TEPCO finally admitted that the hatch was open in April 2012. As usual.
If the steam seen on the top floor of Reactor 3 is from inside the Containment Vessel, I would think the radiation level should be in 1,000s, instead of 100s.