Terminal-Bench 4.0.0 · GPT-5.6 Luna / Codex
PublicRerun job
330 / 330 finished
Started Aug 26, 2026, 6:18 PM · Ran for 14h 30m
315 completed15 timeouts0 errors0 in progress0 pending
Average reward
0.17
Across scored trials
Total tokens
11,565,143,913
35,045,891 tokens/trial · 98% cache hit
Inference cost
$346.89
$1.05 per finished trial
Sandbox cost
Pending
Costs appear after usage is processed
Results
Metric: reward
Group by0 groups
Loading results…