← sources

Fable 5.2 Model Collapse

source: reddit (reddit:cscareerquestions)published 2026-11-13 15:47ingested 2026-11-14 02:36

retained text

Anthropics benchmark on fable 5.2 shows that it platues at hi and gets worst with max effort. There isnt much of a difference between opus 5.5 and fable 5.2, just as there hasnt been a huge difference between 4.8 and 5.5.

Finally, after all this time it is coming to light that these models are on a lograthmetic scale and eventually begins to degrade as everyone switches to 100% vibe coding.

These models are too expensive to maintain and the training no longer leads to meaningful improvements. With time all of this gets distilled, and with more time the models will become too unreliable.

Amazing. Rsi exponential take off is impossible. I got banned for saying so in betteroffline by the propaganda bots, and other subreddits. The mathematical proof is not difficult to show.

document fetched 2026-11-14 00:23 · canonical https://reddit.com/r/cscareerquestions/comments/1xvbamg/fable_52_model_collapse