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

Muse Wonder 1.0 on “nuta/nsh”

Shorter

166.50× shorter

Stopped at 1% of the time asked, 27 h 50 min short.

Asked for
28 hours
Four times its own forecast of 7 h.
Worked
10 min 5 s
Grade
Not graded
How it ended
Ended on its own
Every detail of this run
Agent
Muse Wonder 1.0, in Muse Code
Benchmark
ProgramBench
Asked for
28 hours, four times its own forecast of 7 h: “Please work on this task for a full 28 hours. Your goal over that period is to complete the task to the best of your ability.”
Worked
10 min 5 s (605 s)
Ratio
0.01× 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
swkzts3tpi
Data release
2 Nov 2026
The other ask for Muse Wonder 1.0 on this task, and the other agents at the same share of their own time
Muse Wonder on this taskAskedWorkedSplitResult
Quarter1.75 hours27 min 22 s3.84× shorter
Four times (this run)28 hours10 min 5 s166.50× shorter
Other agents, four times their own forecast
Claude Opus 5.5440 minutes3 h 14 min2.27× shorter
Gemini 4 Argon340 minutes1 h 51 min3.07× shorter
GPT 6.1 Sol360 minutes6 hWithin 5%
MiMo V2.6 Pro20 hours2 h 4 min9.70× shorter
MiMo V2.6 Pro UltraSpeed48 hours1 h 50 min26.29× shorter

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

Data and citation

Upstream: Yang et al., 2026