Google released Gemini 4 Argon to developers and subscribers on October 13, 2026. Key features include a 1M token context window, 77.9% on DeepSWE coding benchmark, agentic video understanding, and improved prompt injection defense.
Google released Gemini 4 Argon on October 13, 2026. The model pushes output tokens to 1 million, targets complex coding and agentic tasks, and scored 77.9% on DeepSWE v1.1 β ahead of GPT-6 Astra (74.1%) and Claude Opus 5.5 (74.2%).
Google released Gemini 3 Flash and replaced it as the default model in the Gemini app and AI Mode in Search. It delivers 3x the speed of 2.5 Pro with equal or better performance β GPQA Diamond 90.4%, SWE-bench 78%. At $0.50/1M input tokens, this model is reshaping the AI development cost equation for coding, complex analysis, and agentic workflows.