aivs.tv

Same model over time

12Single-run anecdote

One model, the same prompt, two dates: 12 posts from 8 people on X, for Claude Opus 5.5 and GPT-6 Astra. Same model, same prompt, run on different dates: one sample each, not a controlled test. Model outputs vary from run to run, so the poster’s claim is not verified here.

What each poster says

Same-model-over-time posts: model, run dates and the poster’s own view
PostModelRunsPoster’s view
Eiffel Tower benchmark, launch vs Oct 3 vs Oct 22 @EnvolDevClaude Opus 5.5at launch vs Oct 3 vs Oct 22They say each run looks worse than the one before, and that the model now burns tokens faster.
Cel-shaded 3D scenery from a photo, a month apart @ishuagra02Claude Opus 5.5Sep 22 vs Oct 21They wouldn’t say Opus 5.5 is nerfed in capability: the rerun cost 80% more but is larger and perhaps better; whether usage drains faster is still open.
Stickman lightsaber fight rerun, Oct 1 vs Oct 20 @DanielZambriniClaude Opus 5.5Oct 1 vs Oct 20They say the Oct 20 rerun is as good as launch, maybe better, and the only big difference is that it used more tokens.
Mentos and cola eruption simulation @ivanainaiClaude Opus 5.5Sep 30 vs Oct 20They say the Oct 20 rerun of the Sep 30 Mentos prompt came out worse, and ask whether Opus 5.5 was quietly nerfed.
Night train benchmark rerun @EnvolDevGPT-6 Astraat launch vs Oct 6They see no nerf: today's GPT-6 Astra night train looks much better than the launch run, with more detail in the train, bridge and rails.
Night train benchmark rerun @EnvolDevClaude Opus 5.5at launch vs Oct 4They say today's Opus 5.5 night train is worse than the launch run: small details up, overall scene down, as in their Eiffel Tower rerun.
Black Mirror X post animation @krika2810Claude Opus 5.5at launch vs Oct 4They say the launch version's quality is clearly better than now and find the change disappointing.
Stained-glass hibiscus @ivanainaiClaude Opus 5.5Sep 30 vs Oct 3They point to visibly different petals from the same prompt on Sep 30 and Oct 3, ask whether Opus 5.5 got nerfed or improved, and don't pick a better run.
Eiffel Tower benchmark rerun @EnvolDevClaude Opus 5.5at launch vs Oct 3They prefer the launch run: the tower may have slightly more detail now, but the buildings, trees and river look worse, and one run doesn't prove a nerf.
Coastal village 3D scene @Mr_SalioClaude Opus 5.5Sep 23 vs Oct 3They say today's output looks noticeably weaker than the launch run, with less detail and polish, and that Opus 5.5 feels a little nerfed.
3D Mario and Mickey Mouse models @luckeyfaradayClaude Opus 5.5Sep 30 vs Oct 3They reran the same Mario and Mickey prompts three days apart and ask whether Opus 5.5 was nerfed, without saying which run is better.
Open-world GTA 6 India game @BuildFastWithAIClaude Opus 5.5at launch vs Oct 2They say the launch-day run is clearly better than today's on the same GTA 6 India prompt, and that Anthropic nerfed Opus 5.5 after launching Sonnet 5.5.