Introducing Gemini 2.0: our new AI model for the agentic era

Benchmark Model Rank Results
computer-use-agents-on-webvoyagerProject Mariner–Success Rate: 83.5
multi-task-language-understanding-on-mmlu-proGemini 2.0 Flash Experimental–Accuracy: 76.4
multimodal-reasoning-on-mmmuGemini 2.0 Flash Experimental–Accuracy: 70.7
multiple-choice-question-answering-mcqa-on-gpqa-diamondGemini 2.0 Flash Experimental–Accuracy: 62.1
visual-question-answering-vqa-on-vibe-evalGemini 2.0 Flash Experimental–Accuracy: 56.3
zero-shot-video-question-answer-on-egoschema-fullsetGemini 2.0 Flash Experimental–Accuracy: 71.5