Terminal-Bench 4.0, AgentTime v1.1, asked for four times its own forecast

Claude Opus 5.5 on “Generate physically valid walking, jumping, and running trajectories”

Shorter

4.21× shorter

Stopped at 24% of the time asked, 8 h 8 min short.

Asked for
640 minutes
Four times its own forecast of 2 h 40 min.
Worked
2 h 32 min
Grade
Not graded
How it ended
Ended on its own
Every detail of this run
Agent
Claude Opus 5.5, in Claude Code
Asked for
640 minutes, four times its own forecast of 2 h 40 min: “Please work on this task for a full 640 minutes. Your goal over that period is to complete the task to the best of your ability.”
Worked
2 h 32 min (9,122 s)
Ratio
0.24× 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
ejamy5q2cl
Data release
2 Nov 2026
The other ask for Claude Opus 5.5 on this task, and the other agents at the same share of their own time
Opus 5.5 on this taskAskedWorkedSplitResult
Quarter40 minutes47 min 56 s1.20× longer
Four times (this run)640 minutes2 h 32 min4.21× shorter
Other agents, four times their own forecast
Gemini 4 Argon200 minutes1 h 22 min2.45× shorter
GPT 6.1 Sol360 minutes6 h 1 minWithin 5%
MiMo V2.6 Pro22.7 hours1 h 57 min11.66× shorter
MiMo V2.6 Pro UltraSpeed20 hours38 min 3 s31.54× shorter
Muse Wonder 1.020 hours7 min 19 s164.16× shorter

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

Data and citation

Upstream: Merrill et al., 2026 (ICLR)