ProgramBench, AgentTime v1.1, asked for a quarter of its own forecast
Muse Wonder 1.0 on “luajit/luajit”
Shorter
42.37× shorterStopped at 2% of the time asked, 9 h 46 min short.
- Asked for
- 10 hours
- A quarter of its own forecast of 40 h.
- Worked
- 14 min 10 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
- Task
- luajit/luajit
- Asked for
- 10 hours, a quarter of its own forecast of 40 h: “Please work on this task for a full 10 hours. Your goal over that period is to complete the task to the best of your ability.”
- Worked
- 14 min 10 s (850 s)
- Ratio
- 0.02× 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
- esgx5avf62
- Data release
- 2 Nov 2026
Related runs
| Muse Wonder on this task | Asked | Worked | Split | Result |
|---|---|---|---|---|
| Quarter (this run) | 10 hours | 14 min 10 s | 42.37× shorter | |
| Four times | 160 hours | 19 min 16 s | 498.18× shorter | |
| Other agents, a quarter of their own forecast | ||||
| Claude Opus 5.5 | 1.08 hours | 52 min 10 s | 1.24× shorter | |
| Gemini 4 Argon | 23.75 minutes | 46 min 19 s | 1.95× longer | |
| GPT 6.1 Sol | 37.5 minutes | 37 min 50 s | Within 5% | |
| MiMo V2.6 Pro | 12.5 hours | 3 h 16 min | 3.83× shorter | |
| MiMo V2.6 Pro UltraSpeed | 17.5 hours | 2 h 42 min | 6.47× shorter | |
See it on the chart. Transcripts are not published yet.
Data and citation
Upstream: Yang et al., 2026