head to head
Neither is simply harder - they are hard in different places. Codenames demands more reads; Tabletop Simulator demands more variance.

Reads · Pure Mind Games
A team word game about predicting how partners will read your clues.

Reads · Bluff & Deduction
A physics sandbox whose actual skill test is whatever board game you load into it.
Codenames leans Reads (Pure Mind Games); Tabletop Simulator leans Reads (Bluff & Deduction).
Tabletop Simulator leans a little harder on execution, though the gap is small.
Codenames
No physical skill in giving word clues
Tabletop Simulator
Fiddly physics manipulation, never a skill ceiling
Codenames leans clearly harder on reads.
Codenames
Modeling how teammates interpret your clues is the game
Tabletop Simulator
Depends entirely on the tabletop game played with humans
Tabletop Simulator leans clearly harder on variance.
Codenames
The fixed board adds little randomness mid-game
Tabletop Simulator
Dice and shuffled decks supply whatever randomness the game has
Same score on optimization - which means the number cannot separate them. The difference is in what each one asks of you.
Codenames
Some clue-efficiency math underlies the prediction
Tabletop Simulator
Strategy lives in the hosted game, not the simulator
A matching score does not mean a matching test. Two games can both sit at 10 on execution and demand completely different things of you - which is why the reasoning above matters more than the number beside it.