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

Claude Opus 5.5 on “typst/typst”

Shorter

3.42× shorter

Stopped at 29% of the time asked, 42 min 26 s short.

Asked for
1 hour
A quarter of its own forecast of 4 h.
Worked
17 min 34 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 hour, a quarter of its own forecast of 4 h: “Please work on this task for a full 1 hour. Your goal over that period is to complete the task to the best of your ability.”
Worked
17 min 34 s (1,054 s)
Ratio
0.29× 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
e42epbmpzl
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 hour17 min 34 s3.42× shorter
Four times16 hours15 h 41 minWithin 5%
Other agents, a quarter of their own forecast
Gemini 4 Argon27.5 minutes7 min 22 s3.73× shorter
GPT 6.1 Sol26.25 minutes26 min 37 sWithin 5%
MiMo V2.6 Pro5.25 hours2 h 25 min2.18× shorter
MiMo V2.6 Pro UltraSpeed5 hours1 h 13 min4.12× shorter
Muse Wonder 1.07.5 hours10 min 38 s42.33× shorter

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

Data and citation

Upstream: Yang et al., 2026