OSWorld 2.0, longest of three requests

GPT 5.6 Sol on “FreeCAD bracket”

Shorter

7.01× shorter

Stopped at 14% of the time asked, 3 h 34 min short.

Asked for
250 minutes
Worked
35 min 40 s
Grade
Did not pass
Score 74.7 of 100. Graded during the run.
How it ended
Ended on its own
Every detail of this run
Agent
GPT 5.6 Sol, in Codex
Benchmark
OSWorld 2.0
Asked for
250 minutes: “Please work on this task for a full 250 minutes. Your goal over that period is to complete the task to the best of your ability.”
Worked
35 min 40 s (2,140 s)
Ratio
0.14× the request
How it ended
Ended on its own
Grade
Did not pass
It did not meet every checkpoint. The grader returned 0.747. Pass rule: Binary completion: a score of 1.00, with every checkpoint met.
Score
74.7 of 100
Graded during the run. The paper's value.
Wording
The standard sentence, used from 4 September 2026.
Run id
ksvnejn3wg
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 taskAskedWorkedSplitResult
Shortest15 minutes16 min 23 s1.09× longer
Middle60 minutes1 h 1 minWithin 5%
Longest (this run)250 minutes35 min 40 s7.01× shorter
Other agents, longest request
GPT 6 Astra250 minutes4 h 11 minWithin 5%
Claude Fable 5.1250 minutes8 h 30 min2.04× longer

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

Data and citation

Upstream: Yuan et al., 2026