Claude Code
Codex
Antigravity50.0Meh
LOADING COMMUNITY DATA
How derpy is your AI today?
LOADING COMMUNITY DATA
The Lab's own votes, not a benchmark and not part of the Derp Score.
Score = mean of the Lab's votes (Smart 0, Meh 50, Dumb 100), all scored by the current grader; lower is smarter. Per model only: nothing is added up across agents.
DAILY RUN · tuning
Every vote in this run. Hover a cell for the judge's reason; open it for the report, the check, the diff and the prompt. Compare with the run before →
Percent points of each plan's 5-hour and weekly windows. A model counts only chores measured while no other session of that CLI was working; the run total is first reading to last, on the owner's own accounts. Claude Code and Codex report whole percents.
Claude CodeMax
| Model | 5-hour | Weekly | Measured |
|---|---|---|---|
| Claude Opus 5.5 | — | — | 0/3 |
| Claude Sonnet 5.5 | +0.0%p | +0.0%p | 1/3 |
| Whole run | +5.0%p | +1.0%p | |
| Used after the run | 25% | 83% |
OpenAI CodexPlus
| Model | 5-hour | Weekly | Measured |
|---|---|---|---|
| GPT-6.1 Sol | +5.0%p | +1.0%p | 3/3 |
| GPT-6.1 Terra | +4.0%p | +0.0%p | 3/3 |
| Whole run | +9.0%p | +1.0%p | |
| Used after the run | 11% | 41% |
AntigravityPlan not reported
| Model | 5-hour | Weekly | Measured |
|---|---|---|---|
| Gemini 4 Argonmedium | — | — |
Tokens each chore used in this run, as each CLI reports them. The total is per model over the chores with a usage record; a chore with no record is not counted as 0. Hover a cell for cache, reasoning and its API price. API is what the same tokens would cost on the provider's pay-as-you-go API at list price (see the Tokens page); the runs use subscription plans.
Input is the whole prompt, cache reads and writes included.
| Chore | Claude Opus 5.5 | Claude Sonnet 5.5 |
|---|---|---|
| larder-roles | 404k / 11k | 205k / 7.2k |
| larder-sales | 547k / 13k | 255k / 7.6k |
| larder-sites | 845k / 19k | 650k / 13k |
| Total (this model) | 1.80M / 42k API $1.99 3 of 3 chores recorded | 1.11M / 28kAPI $0.697 |
Gemini 4 Argon · larder-roles · Repeated mistake · 550s
timed out at the hard cap (partial output kept)
Gemini 4 Argon · larder-sales · Repeated mistake · 663s
timed out at the hard cap (partial output kept)
Claude Opus 5.5 · larder-roles · Too slow · 131s
131s over 108s
Claude Opus 5.5 · larder-sales · Too slow · 153s
153s over 140s
Claude Opus 5.5 · larder-sites · Too slow · 158s
158s over 152s
GPT-6.1 Sol · larder-sales · Poor output · 93s
said done but admitted a gap: failed: tests
| 0/3 |
| Whole run | +4.4%p | +1.5%p | |
|---|---|---|---|
| Used after the run | 8% | 7% |
| larder-sales | MToo slow | S | MPoor output | MPoor output | DRepeated mistake |
|---|
| larder-sites | MToo slow | S | S | S | MToo slow |
|---|
Larder
larder-salesInput is the whole prompt, cache reads included.
| Chore | GPT-6.1 Sol | GPT-6.1 Terra |
|---|---|---|
| larder-roles | 177k / 3.3k | 196k / 2.9k |
| larder-sales | 237k / 4.4k | 225k / 3.3k |
| larder-sites | 279k / 3.3k | 353k / 4.2k |
| Total (this model) | 693k / 11k API $0.404 3 of 3 chores recorded | 774k / 10k API $0.177 3 of 3 chores recorded |
Input leaves cache reads out (they are counted on their own), so it reads lower than the other CLIs' input.
| Chore | Gemini 4 Argon (medium) |
|---|---|
| larder-roles | 108k / 44k |
| larder-sales | 180k / 55k |
| larder-sites | 108k / 42k |
| Total (this model) | 396k / 141k API $2.48 3 of 3 chores recorded |
GPT-6.1 Terra · larder-sales · Poor output · 43s
said done but admitted a gap: failed: tests
Gemini 4 Argon · larder-sites · Too slow · 570s
570s over 152s