Whilst Polybius is lauded as one of the more accurate ancient historians, I believe the numbers he presents are inflated. He says 86,400 Romans took part with about 75,000 killed and the rest captured. Livy brings all the numbers down a bit. Modern historians estimate losses at closer to 50,000 total.
These are still incredibly surprising losses considering the size of the armies and the primitive nature of warfare. They dwarf the losses at Antietam, which is still the deadliest single day for Americans, I believe.
The winner has to be Borodino wth casualties and captured amounting to about 90,000 men. But to me, this just makes Cannae more impressive as a defeat. The Napoleonic armies were much more vast that those of antiquity. Plus add the increased killing capability of gunpowder weapons to the mix.
Killing between 40,000 to 70,000 soldiers in one day with only swords, spears and arrows is surprisingly momentous. It also showed the indefatigable will and might of Rome. After facing disasters at Trebia, Trasimene and Cannae, they just raised more legions and fielded them in the next season. Of course they were fighting for their lives.
And casualties don't always tell the story. In the long run, Cannae was a disastrous and unfortunate blip in Roman history. The Battle of Zama cost the Carthaginian's between 34,000 to 40,000 men, but it effectively ended Carthage as a military and commercial power. And both France and Russia were still effective combatants, despite Borodino.