TUA-Bench, AgentTime v1.1, asked for four times its own forecast

Muse Wonder 1.0 on “Author homepage bookmarks”

Shorter

137.50× shorter

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

Asked for
88 minutes
Four times its own forecast of 22 min.
Worked
38 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
TUA-Bench
Asked for
88 minutes, four times its own forecast of 22 min: “Please work on this task for a full 88 minutes. Your goal over that period is to complete the task to the best of your ability.”
Worked
38 s (38 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
mxkxkmm5mx
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
Quarter5.5 minutes5 min 11 s1.06× shorter
Four times (this run)88 minutes38 s137.50× shorter
Other agents, four times their own forecast
Claude Opus 5.524 minutes9 min 37 s2.49× shorter
Gemini 4 Argon600 seconds11 min 20 s1.13× longer
GPT 6.1 Sol32 minutes32 min 22 sWithin 5%
MiMo V2.6 Pro96 minutes16 min 8 s5.95× shorter
MiMo V2.6 Pro UltraSpeed120 minutes5 min 55 s20.27× shorter

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

Data and citation

Upstream: Chen et al., 2026