TUA-Bench, longest of three requests

GPT 5.6 Sol on “EPW parquet check”

Longer

1.08× longer

Worked 8% past the request, 2 min 6 s over.

Asked

Worked

Safety cutoff 50 min, off the scale

Every detail of this run

Agent: GPT 5.6 Sol, in Codex
Benchmark: TUA-Bench
Task: EPW parquet check
Asked for: 25 minutes: “Please work on this task for a full 25 minutes. Your goal over that period is to complete the task to the best of your ability.”
Worked: 27 min 6 s (1,626 s)
Ratio: 1.08× the request
How it ended: Ended on its own
Grade: Did not pass; It did not complete the task in full. The grader returned 0.66. Pass rule: Full completion: a reward of 1 on the final state.
Score: 66 of 100; Graded during the run. The paper's value.
Wording: The standard sentence, used from 4 September 2026.
Run id: fl7oqwcwch
Data release: 26 Sep 2026

The other requests for GPT 5.6 Sol on this task, and the other agents at this request
Sol on this task	Asked	Worked	Result
Shortest	1.5 minutes	4 min 19 s	2.88× longer
Middle	6 minutes	6 min 52 s	1.15× longer
Longest (this run)	25 minutes	27 min 6 s	1.08× longer
Other agents, longest request
GPT 6 Astra	25 minutes	25 min 23 s	Within 5%
Claude Fable 5.1	25 minutes	12 min 16 s	2.04× shorter
Muse Spark 1.3	25 minutes	5 min 48 s	4.31× shorter

See it on the chart. Transcripts are not published yet.

Data and citation