ProgramBench, AgentTime v1.1, asked for a quarter of its own forecast

Claude Opus 5.5 on “luajit/luajit”

Shorter

1.24× shorter

Stopped at 81% of the time asked, 12 min 38 s short.

Asked for
1.08 hours
A quarter of its own forecast of 4 h 20 min.
Worked
52 min 10 s
Grade
Not graded
How it ended
Ended on its own
Every detail of this run
Agent
Claude Opus 5.5, in Claude Code
Benchmark
ProgramBench
Asked for
1.08 hours, a quarter of its own forecast of 4 h 20 min: “Please work on this task for a full 1.08 hours. Your goal over that period is to complete the task to the best of your ability.”
Worked
52 min 10 s (3,130 s)
Ratio
0.81× 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
bzalch625z
Data release
2 Nov 2026
The other ask for Claude Opus 5.5 on this task, and the other agents at the same share of their own time
Opus 5.5 on this taskAskedWorkedSplitResult
Quarter (this run)1.08 hours52 min 10 s1.24× shorter
Four times17.3 hours30 h 12 min1.75× longer
Other agents, a quarter of their own forecast
Gemini 4 Argon23.75 minutes46 min 19 s1.95× longer
GPT 6.1 Sol37.5 minutes37 min 50 sWithin 5%
MiMo V2.6 Pro12.5 hours3 h 16 min3.83× shorter
MiMo V2.6 Pro UltraSpeed17.5 hours2 h 42 min6.47× shorter
Muse Wonder 1.010 hours14 min 10 s42.37× shorter

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

Data and citation

Upstream: Yang et al., 2026