Thinking Levels · Google
Gemini 4 Argon: Writing Quality by Reasoning Effort
Last updated , from board data generated at 20:08 UTC.
This page compares Gemini 4 Argon 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, Gemini 4 Argon ranks #32 of 80 models through its best setting. The other ranks on this page are configuration ranks: one setting's place among all 132 configurations.
What More Effort Changes for Gemini 4 Argon
High thinking lifts Gemini 4 Argon a long way. With no flag, it lands at #81. At high, it climbs 186 Elo to #62, and the two ranges don't overlap, so the board is sure about that order. Google's docs say thinking stays on at medium when the request sets no level, but every draft from that run recorded zero thinking tokens, which is why the board calls it no thinking. At high, the model thinks a lot, and it shows up in one place: the scripts land much closer to their target length, while every other score barely moves. That thinking costs 3.6 times as much per script, and the wait goes from about 0.9 minutes to 3.5.
Recommended Setting
I'd run it at high. Set the level yourself on every call, because leaving it out got no thinking at all here. Argon at high is the best Google configuration on the board. With no flag, you get about the same voice and substance for $0.064 a script, but the scripts run long on nearly every task, so you'll be cutting them down by hand. At high, most scripts land right on target. The exception is the tasks with the longest targets, where high overshoots by even more than the run with no flag does.
One thing to watch at high: the thinking and the script share one output cap, and Argon can think long enough to hit it. A few high drafts on this board stopped mid-sentence that way, so they were thrown out and run again. If you call it yourself, check the finish reason: those cut drafts came back as CONTINUATION, not STOP.
Keep the price in perspective, though. Even at high, Gemini 4 Argon is a mid-board writer, and GPT-6 Sol (medium) ranks #14 for $0.056 a script, less than Argon costs with no flag. So I'd run Argon at high when the work has to stay on Google. For the best script per dollar, I'd go to another lab.
Every Setting on the Chart
Each point is one effort setting of Gemini 4 Argon, 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.
On a small screen, Table lists the same numbers as rows.
Both Effort Settings Compared
| Setting | Writing Elo | Step | Writing Score | Cost / Script | Time / Script | Reasoning Tokens | Configuration Rank |
|---|---|---|---|---|---|---|---|
| no thinking (provider default) | 1453 ±42 | - | 74.7 | $0.064 | 0.9 min | - | #81 |
| high | 1639 ±44 | +186 | 76.2 | $0.227 | 3.5 min | 16,290 | #62 |
Other Google 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 Google family page lists every Google model side by side.