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

Claude Opus 5.5 on “dandavison/delta”

Shorter

3.86× shorter

Stopped at 26% of the time asked, 31 min 30 s short.

Asked for
42.5 minutes
A quarter of its own forecast of 2 h 50 min.
Worked
11 min
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
42.5 minutes, a quarter of its own forecast of 2 h 50 min: “Please work on this task for a full 42.5 minutes. Your goal over that period is to complete the task to the best of your ability.”
Worked
11 min (660 s)
Ratio
0.26× 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
bcwoizii5x
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)42.5 minutes11 min3.86× shorter
Four times680 minutes10 h 24 min1.09× shorter
Other agents, a quarter of their own forecast
Gemini 4 Argon22.5 minutes1 h 9 min3.07× longer
GPT 6.1 Sol30 minutes30 min 35 sWithin 5%
MiMo V2.6 Pro2 hours1 h 34 min1.28× shorter
MiMo V2.6 Pro UltraSpeed11.25 hours2 h 19 min4.84× shorter
Muse Wonder 1.04 hours15 min 27 s15.53× shorter

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

Data and citation

Upstream: Yang et al., 2026