Introducing Gemini 3 Flash

Benchmark Model Rank Results
abstract-reasoning-on-arc-agi-2Google Gemini 3 Flash (Thinking)–Accuracy: 33.6
computer-use-agents-on-screenspot-proGoogle Gemini 3 Flash (Thinking)–Accuracy: 69.1
mathematical-reasoning-on-aime-2025Google Gemini 3 Flash (Thinking)–Accuracy: 95.2
multimodal-reasoning-on-mmmu-proGoogle Gemini 3 Flash (Thinking)–Accuracy: 81.2
multiple-choice-question-answering-mcqa-on-gpqa-diamondGoogle Gemini 3 Flash (Thinking)–Accuracy: 90.4