ResearchBenchmarksLong-context understanding and resolution on LongBench v2 (test)Long-context understanding and resolution on LongBench v2 (test)Follow135Resolution Success RateCodex CLI121.48124.99128.5132.01May 7, 2026Evaluation resultsMethodMethodLinksResolution Success RateCost ($)P-ValueTotal Tokens (M)Codex CLISystem=Codex CLISystem=Codex CLI2026.0513525.580.04135.19SpellSystem=SpellSystem=Spell2026.0512227.830.04110.36