Terminal-Bench 4.0.0 · GPT-6 Astra / Codex · reasoning effort max
PublicRerun job
330 / 330 finished
Started Aug 31, 2026, 4:00 AM · Ran for 38h 41m
330 completed0 timeouts0 errors0 in progress0 pending
Average reward
0.58
Across scored trials
Total tokens
1,529,778,322
4,635,692 tokens/trial · 96% cache hit
Inference cost
$3,267.18
$9.90 per finished trial
Sandbox cost
—
Results
Metric: reward
Group by0 groups
Loading results…