Gemini 3 Flash Released

Gemini 3 Flash has been released. It outperforms 2.5 Pro in benchmarks while being significantly cheaper. In some benchmarks, it even surpasses Gemini 3 Pro. It supports the same million tokens of context as the larger model. It accepts text, audio, and images as input, but currently only outputs text. Gemini 3 Flash is a hybrid reasoner — reasoning can be disabled, similar to 2.5 Flash. The token price has increased compared to 2.5 Flash, but they claim that the efficiency of token usage has also improved. In theory, this should compensate for the price increase, but tests indicate otherwise, at least with reasoning enabled. The model is already available in ai.studio, Antigravity, chat, and API. The Google AI mode is also transitioning to Gemini 3 Flash today (finally, there will be a proper model there).