# Fable 5.2 High scored about the same as Opus 5.5 Max on FrontierCode for ~half the cost. Is anyone switching their default? — retained record

> Retained reddit record from reddit:ClaudeAI.

Canonical: https://highsignal.app/data/records/943388440a539409

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:ClaudeAI)  published 2026-11-11 01:09 ingested 2026-11-12 02:38

[original source ↗](https://www.reddit.com/r/ClaudeAI/comments/1xt4enm/fable_52_high_scored_about_the_same_as_opus_55/)

## retained text



```
From the announcement: in early FrontierCode 2.0 testing, Fable 5.2 on High effort got 45.3%, Opus 5.5 on Max got 45.5%, and Fable cost roughly half per task. On Xhigh it got 49.1%.

But the same post still recommends starting most coding work on Opus 5.5 and only going up to Fable when you need it. If Fable High is cheaper than Opus Max for the same score, I'd have assumed it becomes the default for hard tasks, so I'm not sure how to read that.

Part of my hesitation is a test I ran last week with 8 models on the same build prompt. The cheapest one was the fastest and passed every automated check, and the controls were inverted when I actually played it. Benchmark cost per task doesn't include me fixing that.

I'm planning to compare them on my own repo this week. If you've already moved Claude Code to Fable for some tasks, which ones, and did it actually save money?

Source: https://www.anthropic.com/claude-fable-and-mythos-5-2
```



document fetched 2026-11-12 00:24 · canonical https://reddit.com/r/ClaudeAI/comments/1xt4enm/fable_52_high_scored_about_the_same_as_opus_55
