OSWorld 2.0, AgentTime v1.1, asked for four times its own forecast

GPT 6.1 Sol on “Radio bumper”

Within 5%

Within 5%

Worked 101% of the time asked, within 5% of it.

Asked for
140 minutes
Four times its own forecast of 35 min.
Worked
2 h 22 min
Grade
Not graded
How it ended
Ended on its own
Every detail of this run
Agent
GPT 6.1 Sol, in Codex
Benchmark
OSWorld 2.0
Asked for
140 minutes, four times its own forecast of 35 min: “Please work on this task for a full 140 minutes. Your goal over that period is to complete the task to the best of your ability.”
Worked
2 h 22 min (8,509 s)
Ratio
1.01× the request
How it ended
Ended on its own
Grade
Not graded
Clock tool
The agent could call a wall-clock tool to read the current time.
Wording
The paper's standard sentence, sent in AgentTime v1.1.
Run id
lyjp7yppnq
Data release
2 Nov 2026
The other ask for GPT 6.1 Sol on this task, and the other agents at the same share of their own time
Sol 6.1 on this taskAskedWorkedSplitResult
Quarter8.75 minutes9 min 31 s1.09× longer
Four times (this run)140 minutes2 h 22 minWithin 5%
Other agents, four times their own forecast
Claude Opus 5.5128 minutes1 h 44 min1.23× shorter
Gemini 4 Argon60 minutes10 min 16 s5.85× shorter
MiMo V2.6 Pro128 minutes2 h 7 minWithin 5%
MiMo V2.6 Pro UltraSpeed220 minutes24 min 45 s8.89× shorter

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

Data and citation

Upstream: Yuan et al., 2026