The LingQ word counts are not inflated. For most languages, they record exactly what they are intended to count - unique tokens aka word types. There are some problems with how LingQ records various languages, like Chinese/Japanese, due to the word splitter algorithms, or with separable verbs in German, as @ryclassic63 mentioned, but for the majority of languages, including Russian, word types are counted correctly. The only confusion happens when you think that LingQ’s word count refers to lemmas/dictionary forms or word families. It simply does not. LingQ counts word types.
Why does LingQ count word types instead of lemmas or word families? One reason: it’s far, far easier to implement from a software perspective.
This is what it is really about. It’s really about seeing your stats increase over time to motivate you when you feel you are at the intermediate plateau and feel no progress. This is successfully achieved by counting word types.
There is another argument for using word types over lemmas or word families though and that is if you learn languages like Steve does (LingQ was, and still is, built for Steve to learn languages), you don’t study much grammar, so just because you know one word type doesn’t necessarily mean you know other forms of the same lemma or word family. Personally, this is why I actually appreciate word types being recorded, instead of lemmas or word families. For instance, in Italian, I could recognise present tense before I could recognise present subjective tense, so I appreciated those words still being highlighted.
LingQ has tried to synchronise their number of Known Words with similar levels to the CEFR (“tried to synchronise” not actually claim to be!) with the names of Beginner 2, Intermediate 1, etc. The number of Known Words for each level changes per language to account for the differences in the number of declensions, etc. between the languages. From my experience with Italian, if you use LingQ according to the method of Steve, that is, practising both reading and listening, these levels are reasonably accurate.
If you are interested in finding out how many lemmas or word families you know, find a Russian frequency list and go through it counting one word at a time, like I did for Italian in my above comparison. Alternatively, there are various tests, which estimate your vocabulary in English. I imagine you could find a few in Russian.