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

Claude Opus 5.5 on “cweill/gotests”

Shorter

2.53× shorter

Stopped at 39% of the time asked, 21 min 11 s short.

Asked for
35 minutes
A quarter of its own forecast of 2 h 20 min.
Worked
13 min 49 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
35 minutes, a quarter of its own forecast of 2 h 20 min: “Please work on this task for a full 35 minutes. Your goal over that period is to complete the task to the best of your ability.”
Worked
13 min 49 s (829 s)
Ratio
0.39× 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
ibiyff5prf
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)35 minutes13 min 49 s2.53× shorter
Four times560 minutes8 h 14 min1.13× shorter
Other agents, a quarter of their own forecast
Gemini 4 Argon16.25 minutes35 min 31 s2.19× longer
GPT 6.1 Sol15 minutes15 min 25 sWithin 5%
MiMo V2.6 Pro1.75 hours1 h 12 min1.46× shorter
MiMo V2.6 Pro UltraSpeed2.5 hours1 h 13 min2.06× shorter
Muse Wonder 1.01.19 hours14 min 32 s4.91× shorter

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

Data and citation

Upstream: Yang et al., 2026