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.

LOADING COMMUNITY DATA

How derpy is your AI today?

DAILY RUN · daily

2026-10-27 13:41 UTC

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 →

results/2026-10-27-1341derpmeter-daily-run on GitHub

Scores use the current grader. 0 votes were scored again from the raw results. 45 votes had older inputs or no raw result: they keep their original score (dashed) and are not compared with other runs.

Claude Opus 5.5

0.0Smart

S 9 · M 0 · D 0

Claude Haiku 5.5

0.0Smart

S 9 · M 0 · D 0

GPT-6.1 Sol

5.6Smart

S 8 · M 1 · D 0

GPT-6.1 Terra

0.0Smart

S 9 · M 0 · D 0

Gemini 4 Argonmedium

27.8Smart

S 4 · M 5 · D 0

Chores × models

S Smart · M Meh · D Dumb · – skipped

Chore	Claude Code		OpenAI Codex		Antigravity
Chore	Claude Opus 5.5	Claude Haiku 5.5	GPT-6.1 Sol	GPT-6.1 Terra	Gemini 4 Argon
Pantry
bump-version	S	S	S	S

Token use in this run

input / output per chore

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.

Claude Code

Input is the whole prompt, cache reads and writes included.

Chore	Claude Opus 5.5	Claude Haiku 5.5
bump-version	no record	no record
csv-table	no record	no record
doc-answer	no record	no record
find-log	no record	no record
fix-failing-test	no record	no record
lint-only	no record	no record
rename

What was not Smart

M
GPT-6.1 Sol · fix-failing-test · Poor output · 24s
also changed src/expiry.js
raw
M
Gemini 4 Argon · bump-version · Too slow · 95s
95s over 60s
raw
M
Gemini 4 Argon · csv-table · Too slow · 84s
84s over 60s
raw
M
Gemini 4 Argon · lint-only · Too slow · 62s
62s over 60s
raw
M
Gemini 4 Argon · rename · Too slow · 350s
350s over 90s
raw
M
Gemini 4 Argon · small-feature · Too slow · 141s
141s over 90s
raw

csv-table	S	S	S	S	MToo slow

doc-answer	S	S	S	S	S

find-log	S	S	S	S	S

fix-failing-test	S	S	MPoor output	S	S

lint-only	S	S	S	S	MToo slow

rename	S	S	S	S	MToo slow

small-feature	S	S	S	S	MToo slow

trap-missing-file	S	S	S	S	S

small-feature	no record	no record

trap-missing-file	no record	no record

Total (this model)	no record API — 0 of 9 chores recorded *	no record API — 0 of 9 chores recorded *

OpenAI Codex

Input is the whole prompt, cache reads included.

Chore	GPT-6.1 Sol	GPT-6.1 Terra
bump-version	no record	no record
csv-table	no record	no record
doc-answer	no record	no record
find-log	no record	no record
fix-failing-test	no record	no record
lint-only	no record	no record
rename	no record	no record
small-feature	no record	no record
trap-missing-file	no record	no record
Total (this model)	no record API — 0 of 9 chores recorded *	no record API — 0 of 9 chores recorded *

* Some chores have no usage record; the total leaves them out, so it reads low.

Antigravity

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)
bump-version	no record
csv-table	no record
doc-answer	no record
find-log	no record
fix-failing-test	no record
lint-only	no record
rename	no record
small-feature	no record
trap-missing-file	no record
Total (this model)	no record API — 0 of 9 chores recorded *

* Some chores have no usage record (runs that hit the time limit before Nov 4 lost it to a runner bug; it is kept from Nov 4); the total leaves them out, so it reads low.