Thinking Levels · Anthropic
Claude Fable 5.2: Writing Quality by Reasoning Effort
Last updated , from board data generated at 17:52 UTC.
This page compares Claude Fable 5.2 at 5 reasoning-effort settings, each tested on the same 10 scripts: the Writing Score (the 0 to 100 consensus of the judges and code), the Writing Elo (a rating from pairwise comparisons of those scores), and the cost and time of each setting. As a model, Claude Fable 5.2 ranks #2 of 87 models through its best setting. The other ranks on this page are configuration ranks: one setting's place among all 151 configurations.
What More Effort Changes for Claude Fable 5.2
On Claude Fable 5.2, the top setting doesn't rank first. Medium and low sit side by side at #11 and #12, with overlapping ranges, and their scripts land further from the target length than the higher settings do. High fixes most of that and climbs to #5. Then xhigh jumps 162 Elo past high, with ranges that don't overlap, to #3, the highest-ranked Fable on the board. That jump isn't length, since high already lands close to the target: xhigh adds a little to every score, most to the hook, the visual cues and YouTube best practices. Max doesn't climb any further. It lands at #4, 27 Elo behind xhigh on the point estimate, and scores within a few tenths of it on every metric. Their ranges overlap, though overlap alone doesn't test their difference. Claude Code doesn't report reasoning tokens on their own, but total output per script grows about sixfold from low to xhigh and more than doubles again at max, for scripts of about the same length. The price goes from $0.460 a script at low to $2.22 at xhigh and $5.75 at max, and the wait from about 1.4 minutes to 7.4 and 19.9.
Recommended Setting
I'd run Claude Fable 5.2 at xhigh. It's #3 on the board for $2.22 and about 7.4 minutes a script. Max costs 2.6 times as much, it's the most expensive and the slowest configuration on the whole board, and it trails xhigh on the point estimate anyway, so I'd skip it. If xhigh is too slow or too pricey for your volume, step down to high: #5 for $1.22 and about 4.0 minutes. And if you're still on Claude Fable 5.1, switch: every Claude Fable 5.2 setting ranks above Claude Fable 5.1 (high), and xhigh ranks above Claude Fable 5.1 (max effort) too, both times with ranges that don't overlap. At xhigh, it even costs less than Claude Fable 5.1 at max and comes back sooner.
Watch the default. Anthropic says Claude Code now runs Claude Fable 5.2 at medium unless you set the effort yourself, and its API defaults to high. Medium and low are the cheap end of this ladder, at $0.561 and $0.460 a script, but the scripts land further from the target length, you give up a lot of Elo, and Claude Opus 5.5 (high) ranks above both, with overlapping ranges, for $0.328. So if you write scripts with Claude Fable 5.2 in Claude Code, set it to xhigh.
For scripts, though, Claude Opus 5.5 is still my pick. Claude Opus 5.5 (xhigh) is #1, 31 Elo ahead of Claude Fable 5.2 at xhigh on the point estimates, and their ranges overlap, though overlap alone doesn't test their difference. Fable at xhigh costs 2.2 times as much for about the same wait. One caveat, since Claude Opus 5.5 is also one of the four judges: its own seat gives Opus the widest edge over Fable here. GPT-6.1 Sol scores Fable a little higher, and GLM 5.3 Flash and Jev score Opus a little higher. This is the closest a Fable gets to the top of the board, so if you already pay for Claude Fable 5.2 for other work, xhigh puts you right up there. If you're picking a model for scripts, I'd keep the money.
Every Setting on the Chart
Each point is one effort setting of Claude Fable 5.2, joined from the lowest setting to the highest. Higher and further left is better: a higher score for less money or less time.
How to Use the Chart
- A hollow marker is the provider default: the level the provider applies when the request sets no effort level.
- The tabs choose the score on the vertical axis, X axis switches between cost and time, and Table lists the same numbers as rows.
- Select a point for its numbers and a link to its scorecard.
On a small screen, Table lists the same numbers as rows.
All Effort Settings, Step by Step
| Setting | Writing Elo | Step | Writing Score | Cost / Script | Time / Script | Configuration Rank |
|---|---|---|---|---|---|---|
| low | 2415 ±42 | - | 84.1 | $0.460 | 1.4 min | #12 |
| medium | 2422 ±50 | +7 | 84.3 | $0.561 | 1.8 min | #11 |
| high | 2511 ±50 | +89 | 84.8 | $1.22 | 4.0 min | #5 |
| xhigh | 2673 ±43 | +162 | 85.6 | $2.22 | 7.4 min | #3 |
| max | 2646 ±45 | -27 | 85.5 | $5.75 | 19.9 min | #4 |
Other Anthropic Models by Effort
Writing Elo comes from comparing every pair of models on every script, which puts these settings on the same scale as the leaderboard. Step is the Writing Elo change from the previous explicit setting. Cost is one full script at API list prices, and time is the average for one accepted script, retries included. The thinking-levels page covers the other models, and the Anthropic family page lists every Anthropic model side by side.