Google began rolling out Gemini 3 Flash, a more efficient version of its latest AI model, approximately one month after launching Gemini 3 Pro in November. The new model delivers pro-grade reasoning performance comparable to the flagship at reduced cost, with benchmarks showing it surpasses prior generations including Gemini 2.5 Pro and competes directly with OpenAI’s GPT-5.2.
The rollout positions Gemini 3 Flash for widespread everyday applications due to its balance of capability and efficiency. Google states that this system maintains high-level reasoning while operating at a fraction of the computational expense of the Pro variant, enabling broader accessibility without performance trade-offs in standard tasks.
Performance evaluations confirm Gemini 3 Flash exceeds Google’s previous models across multiple metrics. It demonstrates substantial improvements over Gemini 2.5 Pro, reflecting advancements in model architecture and training processes that enhance speed and accuracy simultaneously.
In the Humanity’s Last Exam benchmark suite, which tests advanced reasoning under constrained conditions, Gemini 3 Flash achieved a score less than one percentage point below GPT-5.2. This comparison occurred without access to external tools such as web search, isolating the models’ intrinsic capabilities.
Gemini 3 Flash also outperformed GPT-5.2 in select evaluations focused on multimodal processing. The MMMU-Pro benchmark, designed to assess a model’s understanding and reasoning across text, images, and other data types, recorded 81.2 percent for Gemini 3 Flash against 79.5 percent for GPT-5.2.
Google integrates Gemini 3 Flash as the default model in the Gemini App and AI Mode within Search. This deployment extends to all users globally. “That means all of our Gemini users globally will get access to the Gemini 3 experience at no cost, giving their everyday tasks a major upgrade,” Google explains.
Within AI Mode, US users gain direct access to the Nano Banana Pro image generator via the chatbot interface. The process involves selecting “Thinking with 3 Pro” from the model picker, followed by “Create Images Pro.”
Gemini 3 Flash is a game changer because it collapses the traditional trade-off between top-tier reasoning and real-world efficiency. Google has effectively packaged near-flagship intelligence into a model that runs at a fraction of the cost, delivering reasoning performance on par with Gemini 3 Pro and competitive with OpenAI’s GPT-5.2, while being cheap enough to deploy everywhere by default.
Benchmarks back this up: Gemini 3 Flash outperforms earlier Google models like Gemini 2.5 Pro, lands within a single percentage point of GPT-5.2 on Humanity’s Last Exam without tools, and even surpasses it on multimodal reasoning in MMMU-Pro. The real shift, however, is distribution. By making Gemini 3 Flash the default model across the Gemini app and Search’s AI Mode, Google is turning “pro-level AI” into a baseline experience for billions of users, reframing advanced reasoning not as a premium feature, but as everyday infrastructure.
