ResearchBenchmarksTerminal task resolution on Terminal-Bench 1.1 (test)Terminal task resolution on Terminal-Bench 1.1 (test)Follow39Resolution RateCodex CLI35.8836.6937.538.31May 7, 2026Evaluation resultsMethodMethodLinksResolution RateOperational CostProbability PTotal Tokens (M)Codex CLISystem=Codex CLISystem=Codex CLI2026.053946.960.54951.14SpellSystem=SpellSystem=Spell2026.053625.720.5496.96