3 ms·
But actual progress seems to be slower. These modes are releasing more often but aren’t big leaps.
by rubicon33 8mo ago
But actual progress seems to be slower. These modes are releasing more often but aren’t big leaps.
- wahnfrieden 8mo agoGPT 5.3 (/Codex) was a huge leap over 5.2 for coding
- rubicon33 8mo agoEh, sure, but marginally better if not the same as Claude 4.6, which itself was a small bump over Claud 4.5
- gallerdude 8mo agoWe used to get one annual release which was 2x as good, now we get quarterly releases which are 25% better. So annually, we’re now at 2.4x better.
- minimaxir 8mo agoDue to the increasing difficulty of scaling up training, it appears the gains are instead being achieved through better model training which appears to be working well for everyone.