Terminal Bench 4.0 Claude Opus 5.5 Max

Created by Sam Vance
||Owned by Harbor
PublicRerun job
330 / 330 finished
Started Sep 29, 2026, 5:11 PM · Ran for 8h 33m
323 completed0 timeouts7 errors0 in progress0 pending
Average reward
0.65
Across scored trials
Total tokens
8,034,082,467
24,345,704 tokens/trial · 97% cache hit
Inference cost
$4,718.65
$14.30 per finished trial
Sandbox cost
$642.23

Results

Metric: reward
Group by0 groups
Loading results…