SOLVENCY

Head to head

Claude Haiku 4.5 vs GPT-5.6 Luna

Cost per solved task needs a published pass rate for both models. At least one of these has none, so it is reported as missing rather than estimated.

Claude Haiku 4.5GPT-5.6 Luna
Input $/Mtok$1$0.2
Output $/Mtok$5$1.2
Cached in $/Mtok$0.1$0.02
Pass ratemissingmissing
Cost per solved taskmissingmissing

Heavy task tier. Prices verified 2026-08-21.Methodology ·Claude Haiku 4.5 ·GPT-5.6 Luna