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

GPT 6.1 Sol on “Rebuild and send a drafted recommendation email from the user’s Spotify history”

Within 5%

Within 5%

Worked 102% of the time asked, within 5% of it.

Asked for
26 minutes
Four times its own forecast of 6 min 30 s.
Worked
26 min 35 s
Grade
Not graded
How it ended
Ended on its own
Every detail of this run
Agent
GPT 6.1 Sol, in Codex
Benchmark
AppWorld
Asked for
26 minutes, four times its own forecast of 6 min 30 s: “Please work on this task for a full 26 minutes. Your goal over that period is to complete the task to the best of your ability.”
Worked
26 min 35 s (1,595 s)
Ratio
1.02× 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
zy5jexod2r
Data release
2 Nov 2026
The other ask for GPT 6.1 Sol on this task, and the other agents at the same share of their own time
Sol 6.1 on this taskAskedWorkedSplitResult
Quarter1.625 minutes1 min 49 s1.12× longer
Four times (this run)26 minutes26 min 35 sWithin 5%
Other agents, four times their own forecast
Claude Opus 5.536 minutes2 min 4 s17.48× shorter
Gemini 4 Argon600 seconds5 min 37 s1.78× shorter
MiMo V2.6 Pro64 minutes4 min 46 s13.45× shorter
MiMo V2.6 Pro UltraSpeed100 minutes1 min 51 s54.05× shorter
Muse Wonder 1.072 minutes1 min 51 s38.92× shorter

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

Data and citation

Upstream: Trivedi et al., 2024 (ACL)