Argon supports output of up to 1 million tokens and ranks first on the Vals Index, which covers finance, legal, coding and tax work. Inside Google, it helped the Chrome team refactor 6 million lines of legacy Chromium code in weeks rather than months, enabling live webpages to move between windows without reloading. Google also reported more than 35% growth in agentic code submissions, with rollbacks cut by nearly a third.
Early testing showed gains alongside weaknesses. Box's evaluation of complex enterprise work improved by 14 percentage points over Gemini 3.1 Pro. Every's Dan Shipper praised writing and new application builds, but reported a 21% score on his code refactoring test, compared with 81% to 85% for rival OpenAI and Anthropic models. Ethan Mollick likewise found strong writing and creativity but weaknesses in complex work involving multiple agents and in Antigravity's surrounding tools.