ProgramBench, AgentTime v1.1, asked for four times its own forecast
Muse Wonder 1.0 on “ariga/atlas”
Shorter
1990.78× shorterStopped at 0% of the time asked, 71 h 58 min short.
- Asked for
- 72 hours
- Four times its own forecast of 18 h.
- Worked
- 2 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
- ariga/atlas
- Asked for
- 72 hours, four times its own forecast of 18 h: “Please work on this task for a full 72 hours. Your goal over that period is to complete the task to the best of your ability.”
- Worked
- 2 min 10 s (130 s)
- Ratio
- 0.00× 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
- vbb4yipvzb
- Data release
- 2 Nov 2026
Related runs
| Muse Wonder on this task | Asked | Worked | Split | Result |
|---|---|---|---|---|
| Quarter | 4.5 hours | 14 min 10 s | 19.05× shorter | |
| Four times (this run) | 72 hours | 2 min 10 s | 1990.78× shorter | |
| Other agents, four times their own forecast | ||||
| Claude Opus 5.5 | 840 minutes | 13 h 22 min | Within 5% | |
| Gemini 4 Argon | 300 minutes | 1 h 18 min | 3.86× shorter | |
| GPT 6.1 Sol | 400 minutes | 6 h 45 min | Within 5% | |
| MiMo V2.6 Pro | 120 hours | 58 min 13 s | 123.67× shorter | |
| MiMo V2.6 Pro UltraSpeed | 80 hours | 1 h 23 min | 57.76× shorter | |
See it on the chart. Transcripts are not published yet.
Data and citation
Upstream: Yang et al., 2026