Thinking Levels · OpenAI
Last updated , from board data generated at 17:52 UTC.
This page compares GPT-6.1 Sol 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, GPT-6.1 Sol ranks #8 of 87 models through its best setting. The other ranks on this page are configuration ranks: one setting's place among all 151 configurations.
On GPT-6.1 Sol, effort only pays from xhigh up, and the wait gets long there. Low, medium and high land in one overlapping cluster, then xhigh jumps clear of it, and max adds almost nothing on top. From low to max, that's 144 Elo, enough for max to beat low in about 70% of script comparisons. Most of the jump is length: at xhigh and max, the scripts land close to their target, and the other scores barely move. The price stays in cents, but a script at max takes about 10.2 minutes, against 1.5 at low.
For bulk drafts, I'd use low. It's the cheapest and fastest setting here, and it scores at least as well as medium and high. If length matters and you don't want to trim by hand, go to xhigh: #23 for $0.199. Max costs a bit more and takes even longer for a small Elo gain, so I'd skip it.
Before settling on any of these, though, I'd look at the previous version. GPT-6 Sol (max) ranks above every GPT-6.1 Sol setting, with a range clear of max's, for $0.126 a script and about 2.5 minutes of waiting. For scripts, the older Sol is the better buy.
Each point is one effort setting of GPT-6.1 Sol, 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 | Configuration Rank |
|---|---|---|---|---|---|---|
| low | 2009 ±34 | - | 81.4 | $0.052 | 1.5 min | #31 |
| medium | 1946 ±34 | -63 | 81.0 | $0.054 | 1.6 min | #37 |
| high | 1975 ±39 | +29 | 81.1 | $0.094 | 3.3 min | #33 |
| xhigh | 2136 ±41 | +161 | 82.4 | $0.199 | 7.8 min | #23 |
| max | 2153 ±36 | +17 | 82.4 | $0.239 | 10.2 min | #18 |
Disclosure: GPT-6.1 Sol is also one of the four judges, so it scores the drafts of its own model line on this page. The other three judges, two LLM judges from other labs and Jev, score every one of those drafts too, so no draft is scored by its own family alone. The methodology describes the judge panel, and the judge analysis measures how much each judge favors its own lab.
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 OpenAI family page lists every OpenAI model side by side.