Overview
- Gemini 3 Flash is now enabled globally and replaces 2.5 Flash as the free default in the Gemini app.
- Google says the model is beginning a progressive rollout as the base for Search’s AI Mode worldwide.
- According to company benchmarks, the model scores 90.4% on GPQA Diamond, 33.7% on Humanity’s Last Exam, and 81.2% on MMMU‑Pro versus 79.5% for GPT‑5.2.
- Google reports it uses about 30% fewer tokens than Gemini 2.5 Pro for everyday tasks and offers “Fast” for quick answers and “Thinking” for more complex problems.
- The model supports text, image, audio, and video inputs for tasks like analyzing short clips or audio recordings, and Google lists pricing at $0.50 per 1M input tokens and $3 per 1M output tokens.