Towards AITowards AIToneBench

Thinking Levels · DeepSeek

DeepSeek V4 Pro: Writing Quality by Reasoning Effort

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

This page compares DeepSeek V4 Pro at 2 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, DeepSeek V4 Pro ranks #51 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 DeepSeek V4 Pro

More effort doesn't help DeepSeek V4 Pro, and if anything it slips. Asking for xhigh lands 33 Elo below the run with no effort flag, which the board records as high, and the two ranges overlap; that alone doesn't settle their order. You still pay for it: 1.9 times the cost per script, and 3.2 minutes of waiting instead of 1.3.

Recommended Setting

Leave the flag off. The extra effort does pull the scripts closer to the target length, but tone, craft, substance and hook all slip a little, so the judges end up in about the same place. That stands out next to DeepSeek V4 Pro 0813 (max), where turning effort up is worth 588 Elo. This version just doesn't seem to get much out of more thinking.

For scripts, I'd skip it at either setting. DeepSeek V4.1 Flash (max) ranks #36 for $0.017 a script, cheaper than either DeepSeek V4 Pro run.

Every Setting on the Chart

Each point is one effort setting of DeepSeek V4 Pro, 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.

Both Effort Settings Compared

Score colors:82 and up65 to 81.9under 65
SettingWriting EloStepWriting ScoreCost / ScriptTime / ScriptReasoning TokensConfiguration Rank
high (provider default)1299 ±48-73.6$0.0371.3 min3,986#104
xhigh1266 ±53-3373.1$0.0703.2 min15,278#106

The 95% Writing Elo ranges overlap: DeepSeek V4 Pro (high) runs from 1246 to 1343, and DeepSeek V4 Pro (xhigh) from 1213 to 1320.

Other DeepSeek 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 provider default. 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 DeepSeek family page lists every DeepSeek model side by side.