This seems to be a point of some contention to me. It's obvious that a shit tonne of people died in the Gulags, but at the same time the more "extravagant" numbers tossed around seem to originate from Western critics, who obviously had a motive to attack the USSR.
From the official statistics from the Soviets, the death rates varied from 3-20%(With 20% being at the height of the German invasion) - though similarly to their Western critics, they would have plenty of reason to fudge the numbers, even in the period of De-Stalinisation.
Do we know how accurate estimates of the death tolls are? Is there a broad consensus amongst the global community? Or are the statistics just lost to time?
This is a very controversial area that's been fought repeatedly over the past two decades, as part of the wider struggle over revisionism in Soviet studies following the end of the Cold War. Some of the writing on this a decade ago (particularly between Conquest and Wheatcroft) was almost polemical. It's been a few years since I seriously tuned in to the academic community on the issue but the impression I get, from conversations, is that there's increasing acceptance in using the archive figures as a base estimate.
That said, no one (not even Wheatcroft) has argued that the statistics are complete or infallible. They simply provide a starting point. For example, Ellman (Soviet Repression Statistics) points out that it was at times standard practice to release prisoners who were terminally ill and on the point of death, thus artificially reducing the recorded mortality rate. Even this was not consistent practice across the timeline in question, with conflicting orders reigning at different times. So treat the figures as rough guides and minimum figures only.
(Keep in mind though that these statistics were culled from internal NKVD reports. They were not intended for general publication. If they contain distortions, and they do, then it's because of pressures and omissions at camp level, not a deliberate attempt at propaganda.)
So the statistics are not perfect. However, IMO they do provide a much better base than the previous guesswork that historians had to work with. Importantly, they've substantially changed the overall picture of the labour camps. There was much more movement/flow between the camps and the outside world than a previous generation of historians had assumed. Instead of being a one-way ticket to oblivion, it was much more common for people to do shorter stints (still 3-5 years) in labour colonies before being released. The overall trends (eg mortality peaking during the war years) are also much clearer.
Personally, I don't think the old estimates (in which the camps were nothing short of a demographic disaster) are sustainable in the light of archive evidence but do take any statistics with a pinch of salt. They're far from perfect but have significantly furthered our knowledge of Soviet repression. It's still somewhat controversial but I recommend that they be considered a solid starting point. In terms of new revelations from the archives, I suspect that that's as good as we'll ever get.
(Note: I've restricted myself to just mentioning Ellman's paper above because it provides a good and neutral overview of the controversy and the key figures. Let me know if you want more.)