Google's Gemini 4 Argon has outperformed OpenAI's GPT-6 Astra in a benchmark comparison, winning 14 out of 19 tests. While Gemini 4 Argon excelled in areas like knowledge work and long context tasks, particularly with a million-token window, GPT-6 Astra demonstrated superiority in terminal-based tests such as hard engineering and terminal science. The cybersecurity benchmark resulted in a tie, indicating that the competition between these advanced models remains close in certain domains. AI
IMPACT Sets a new benchmark for LLM capabilities in knowledge work and long context, potentially influencing future model development and enterprise adoption.
RANK_REASON Benchmark comparison of two AI models. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →