Towards AITowards AIToneBench

Thinking Levels · Anthropic

Claude Sonnet 5.5: Writing Quality by Reasoning Effort

Last updated , from board data generated at 17:52 UTC.

This page compares Claude Sonnet 5.5 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 Sonnet 5.5 ranks #3 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 Sonnet 5.5

Effort matters enormously on Claude Sonnet 5.5, and every step past medium is a clear gain. It starts at #66 on the board at low and climbs to #6 at max, a 766 Elo swing that would have max beating low on about 99% of scripts. High, xhigh and max each clear the range of the setting below them, so the board is confident about each of those steps. The catch is the bill: max costs 22.7 times what low does, and a script takes about 14.3 minutes instead of 0.7.

Recommended Setting

On Claude Sonnet 5.5 itself, skip low and medium. They sit in one overlapping cluster, and high is already a clear step above them for $0.092 a script. With more budget, xhigh is the sweet spot: #15 on the board for $0.302 and about 3.1 minutes.

The setting that makes this model look best is the one I'd skip. Max is #6 on the board, but Claude Opus 5.5 (high) lands in the same Elo range for $0.328 instead of $1.38, and finishes in 2.1 minutes instead of 14.3. That same Opus setting also beats Sonnet's xhigh, with ranges that don't overlap, for about the same price. So if you can pick the model, I'd pick Claude Opus 5.5 and keep Sonnet at high for cheap, fast drafts.

Every Setting on the Chart

Each point is one effort setting of Claude Sonnet 5.5, 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.
Score on the vertical axis
View
X axis

On a small screen, Table lists the same numbers as rows.

All Effort Settings, Step by Step

Score colors:82 and up65 to 81.9under 65
SettingWriting EloStepWriting ScoreCost / ScriptTime / ScriptConfiguration Rank
low1730 ±35-78.9$0.0610.7 min#66
medium1776 ±41+4679.4$0.0650.6 min#62
high1956 ±44+18081.0$0.0921.0 min#35
xhigh2262 ±42+30683.2$0.3023.1 min#15
max2496 ±52+23484.6$1.3814.3 min#6

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.