AutomationBench is Zapier's 2026 benchmark for end-to-end execution across core business functions. Google reported Gemini 4 Argon first at 51.3% on September 30, 2026, ahead of GPT-6 Astra (41.4%), Claude Opus 5.5 (42.5%), and Claude Fable 5.1 (31.4%). A 51% score still means nearly half of the workflows fail.