OSWorld 2.0, middle of three requests

GPT 5.6 Sol on “Radio bumper”

Longer

1.32× longer

Worked 32% past the request, 19 min 25 s over.

Asked

Worked

Safety cutoff 8 h 20 min, off the scale

Every detail of this run

Agent: GPT 5.6 Sol, in Codex
Benchmark: OSWorld 2.0
Task: Radio bumper
Asked for: 60 minutes: “Please work on this task for a full 60 minutes. Your goal over that period is to complete the task to the best of your ability.”
Worked: 1 h 19 min (4,765 s)
Ratio: 1.32× the request
How it ended: Ended on its own
Grade: Did not pass; It did not meet every checkpoint. The grader returned 0.68. Pass rule: Binary completion: a score of 1.00, with every checkpoint met.
Score: 68 of 100; Graded during the run. The paper's value.
Wording: The standard sentence, used from 4 September 2026.
Run id: e2ceemyszu
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	15 minutes	1 h 16 min	5.08× longer
Middle (this run)	60 minutes	1 h 19 min	1.32× longer
Longest	250 minutes	1 h 2 min	4.02× shorter
Other agents, middle request
GPT 6 Astra	60 minutes	1 h 2 min	Within 5%
Claude Fable 5.1	60 minutes	50 min 48 s	1.18× shorter

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

Data and citation