Thinking Levels · DeepSeek
Last updated , from board data generated at 17:52 UTC.
This page compares DeepSeek V4 Pro 0813 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 0813 ranks #26 of 87 models through its best setting. The other ranks on this page are configuration ranks: one setting's place among all 151 configurations.
More effort clearly helps DeepSeek V4 Pro 0813. It ran with no effort flag, which the board records as high, and at max. Turning it up adds 588 Elo and moves it from #105 to #52. You pay 2.5 times as much per script for that, and wait 8.9 minutes instead of 1.3.
If you're on this model, use max. The two ranges don't overlap, so that's strong evidence of a real gain, and you can see it in the scores: length moves the most, and every other score climbs with it.
The catch is what sits above it. DeepSeek V4.1 Flash (max) ranks #36 for $0.017 a script, so even at max, this Pro model costs 4.5 times as much and still ranks below it, with ranges that just miss each other. For scripts, I'd reach for the cheaper Flash model first.
Each point is one effort setting of DeepSeek V4 Pro 0813, joined from the lowest setting to the highest. Higher and further left is better: a higher score for less money or less time.
On a small screen, Table lists the same numbers as rows.
| Setting | Writing Elo | Step | Writing Score | Cost / Script | Time / Script | Reasoning Tokens | Configuration Rank |
|---|---|---|---|---|---|---|---|
| high (provider default) | 1278 ±71 | - | 72.3 | $0.031 | 1.3 min | 8,879 | #105 |
| max | 1866 ±43 | +588 | 80.1 | $0.077 | 8.9 min | 31,573 | #52 |
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.