Thinking levels · DeepSeek
Does DeepSeek V4 Flash write better when it thinks harder?
Barely. The xhigh setting lands 49 Elo above the default, a gap that sits inside the noise. Money isn't the issue at these prices. The wait is: 3.0 minutes a script at xhigh, against 0.7 minutes on default.
Both settings we ran
| Setting | Elo | Step | Overall | Cost / script | Time / script | Reasoning tokens | Board rank |
|---|---|---|---|---|---|---|---|
| xhigh | 1199 ±46 | - | 77.0 | $0.0033 | 3.0 min | 10,633 | #127 |
| default (no effort flag) | 1150 ±40 | - | 75.8 | $0.0014 | 0.7 min | - | #135 |
The 95% Elo ranges overlap: DeepSeek V4 Flash (default) runs from 1108 to 1188, and DeepSeek V4 Flash (xhigh) from 1157 to 1249.
Which setting should you use?
I'd keep DeepSeek V4 Flash on default. Sitting through that longer wait for a gap this small isn't a trade I'd make, however cheap the script stays. Better yet, I'd move up a version. DeepSeek V4.1 Flash (default), its successor, ranks #44 for $0.0077 a script, and you get that without touching a thinking setting.
Other DeepSeek ladders
Elo comes from comparing every pair of models on every script across the full board, so these settings sit on the same scale as the leaderboard. Cost is one full script at API list prices, and time is one accepted script, retries included. Compare this ladder with other models on the thinking-levels page, or see every DeepSeek model side by side.