← sources

Fable 5.2 Low vs Opus 5.5 Med | Pelican Test | "Go all out"

source: reddit (reddit:ClaudeAI)published 2026-11-10 18:22ingested 2026-11-11 02:37

retained text

Same prompt for both, ending with "go all out".

|Metric|Fable 5.2 Low|Opus 5.5 Med|
|:-|:-|:-|
|Lines|594|1,085|
|File size|40 KB|68 KB|
|Time|14 min 12 s|29 min 30 s|
|Output tokens|51,958|116,722|
|Thinking|23,592 (45%)|70,250 (60%)|
|API-equivalent cost|$3.71|$3.87|

**Fable 5.2 Low:** clean. I couldn't find an obvious mistake. But it's one sunset scene, one interaction (click the pelican), no sound.

**Opus 5.5 Med:** way more ambitious. Sound it synthesized itself (bell, splashes, waves), a full day/night cycle with lamps, headlight and fireflies, click the sea for jumping fish, keyboard shortcuts. But it has visible mistakes: 2 sets of wings (on the handle and on the side of the pelican), the beak on top of the eyes, and the scarf goes through the pelican.

What stood out to me: Opus took 2x the time and 2x the tokens for only 4% more cost. So Fable did half the work for the same money, but the half it did is correct.

document fetched 2026-11-11 00:21 · canonical https://reddit.com/r/ClaudeAI/comments/1xsufmg/fable_52_low_vs_opus_55_med_pelican_test_go_all