Skip to content
Hbench
hbench

humaneval-plus@1.0 (unreviewed) on GPT-5.6 Terra

164 tasks, 1 attempt, open network. Updated 24 Oct 2026.

164 tasks from humaneval-plus@1.0 on ghcr.io. Nobody has reviewed this dataset, so a score here says nothing about the benchmark it's named after.

Build a run for this board

Value

What a task cost each configuration. Claude Code used 36% fewer tokens than Codex, and 1% more time per task.

$ per solved task

Lower is cheaper. Dollars per task divided by the pass rate.

Embed this chart

Harness$ per solved task
Codex0.164.0 · Reference$0.03$0.03 a task
Claude Code2.1.309$0.07$0.07 a task

Tokens per task

Lower is leaner. Input plus output, cached input included.

Embed this chart

HarnessTokens per task
Claude Code2.1.30934k×0.64
Codex0.164.0 · Reference53k×1.00

Agent time per task

Lower is faster. Agent start to finish; setup and tests left out.

Embed this chart

HarnessTime per task
Codex0.164.0 · Reference26 s×1.00
Claude Code2.1.30926 s×1.01

Pass rate against tokens per task

Up and to the left is better. Pass-rate axis starts at 85%.

Embed this chart

85%90%95%100%30k40k50k60kTokens per task, input and outputMore tokens for a lower pass rate than a row on the frontierCodex0.164.0 · 53k · 95.7%Claude Code2.1.309 · 34k · 93.9%
  • On the frontier: no other row has a higher pass rate for less tokens
  • Off the frontier
  • 95% interval of the pass rate
On the frontier, least tokens first: Claude Code 2.1.309, 34k tokens a task, 93.9%; Codex 0.164.0, 53k tokens a task, 95.7%.
  • 2 runs
  • Updated 24 Oct 2026
  • Method