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

Claude Opus 5.5 on “stranger6667/jsonschema”

Within 5%

Within 5%

Worked 97% of the time asked, within 5% of it.

Asked for
680 minutes
Four times its own forecast of 2 h 50 min.
Worked
10 h 56 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
680 minutes, four times its own forecast of 2 h 50 min: “Please work on this task for a full 680 minutes. Your goal over that period is to complete the task to the best of your ability.”
Worked
10 h 56 min (39,376 s)
Ratio
0.97× 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
dxpasfs33e
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
Quarter42.5 minutes33 min 31 s1.27× shorter
Four times (this run)680 minutes10 h 56 minWithin 5%
Other agents, four times their own forecast
Gemini 4 Argon360 minutes1 h 48 min3.34× shorter
GPT 6.1 Sol240 minutes4 hWithin 5%
MiMo V2.6 Pro52 hours1 h 56 min26.83× shorter
MiMo V2.6 Pro UltraSpeed28 hours1 h 15 min22.35× shorter
Muse Wonder 1.044 hours16 min 31 s159.90× shorter

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

Data and citation

Upstream: Yang et al., 2026