Thinking Levels · DeepSeek
Last updated , from board data generated at 17:52 UTC.
This page compares DeepSeek V4.1 Flash at 4 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 V4.1 Flash ranks #17 of 87 models through its best setting. The other ranks on this page are configuration ranks: one setting's place among all 151 configurations.
Thinking lifts DeepSeek V4.1 Flash a long way, and the usual reason to hold back barely applies. Going from no thinking to max adds 394 Elo, enough that max would win about 91% of its script comparisons against the no-thinking run. Even at max, a script costs $0.017 and comes back in about 2.4 minutes.
I'd run it at max. It's #36 on the whole board, and even at 4.6 times the no-thinking price, the bill stays tiny. Most of the climb comes early. Switching thinking on at low is the biggest single step, and going on to high adds another 98 Elo. Neither of those steps has overlapping ranges, though the low and high ranges only just miss. Max adds another 74, with ranges that overlap, and overlap alone doesn't establish whether that last step improves the scripts. So if you want scripts faster, high gets you most of the way in 1.1 minutes.
You can see where the thinking goes in the length scores: each step up lands closer to the target length, and the visual cues get better too. I wouldn't turn thinking off, though. It's the cheapest and fastest run on this ladder, and also the weakest by far.
If you're picking today, I'd take the newer snapshot instead. DeepSeek V4.1 Flash 1017 (max) is #30, costs less than this model does at max and comes back faster. DeepSeek V4.1 Flash 1017 (high) also ranks above this model at high, with ranges that don't overlap.
Each point is one effort setting of DeepSeek V4.1 Flash, 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 |
|---|---|---|---|---|---|---|---|
| none | 1562 ±45 | - | 76.8 | $0.0037 | 0.3 min | - | #82 |
| low | 1784 ±57 | +222 | 79.3 | $0.0077 | 0.7 min | 6,740 | #60 |
| high | 1882 ±39 | +98 | 80.3 | $0.012 | 1.1 min | 13,809 | #48 |
| max | 1956 ±40 | +74 | 81.0 | $0.017 | 2.4 min | 22,010 | #36 |
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 DeepSeek family page lists every DeepSeek model side by side.