Gemini 3 Flash: Google has stepped up its AI rollout with the release of Gemini 3 Flash, a faster and more efficient model designed for everyday use. The launch follows the recent debut of Gemini 3 Pro and marks a wider push to bring the latest generation of Gemini models across Google’s products, including the Gemini app, Search, and developer platforms.
The new Flash model is available immediately through the Gemini API, Vertex AI, AI Studio, and Antigravity, while also becoming the default experience for many users.
Faster Performance, Stronger Benchmarks
According to Google, Gemini 3 Flash delivers clear gains over its predecessor, Gemini 2.5 Flash, particularly in reasoning, general knowledge, and advanced domain tests. The company says the model shows modest improvements in standard academic benchmarks and a much larger jump in complex evaluations such as Humanity’s Last Exam (HLE), where it scored 33.7% without tool use, placing it close to Gemini 3 Pro.
Google also highlights stronger results in coding-related tests. In the widely used SWE-Bench Verified, Gemini 3 Flash recorded a significant improvement compared to the earlier Flash model, narrowing the gap with Pro-tier models. General knowledge accuracy has also improved sharply, making the new model less prone to factual errors.
Closer To Pro Models, At Lower Cost
While Gemini 3 Pro remains Google’s most capable offering, Gemini 3 Flash now performs much closer to the Pro tier while running workloads up to three times faster, the company says. It also retains advanced features such as interactive simulations and multimodal content generation.
For developers, the pricing reflects its middle-ground position. Gemini 3 Flash costs $0.50 per million input tokens and $3 per million output tokens, making it more affordable than Pro models, though slightly pricier than the previous Flash generation. By comparison, Gemini 3 Pro is priced at $2 per million input tokens and $12 per million output tokens.
Default Model In Gemini App And AI Search
To simplify model selection, Google is making Gemini 3 Flash the default model in the Gemini app and web interface. It will power both the “Fast” setting and the “Thinking” mode, which uses simulated reasoning to improve responses. Gemini 3 Pro will continue to be available as an option for users who want higher capability.
The Flash model is also rolling out to Google Search’s AI Mode, where it will become the default engine. This means free users are expected to see noticeable performance improvements when interacting with AI-powered search features. AI Overviews will continue to dynamically choose models based on task complexity, with faster models likely handling simpler queries.
Google confirmed that Gemini 3 Pro and Gemini 3 Pro Image will also expand within AI Mode, particularly for US-based users. While free access will have usage limits, Pro and Ultra subscribers will receive significantly higher allowances.
With Gemini 3 Flash, Google is positioning a faster, more capable AI model as the new baseline—aiming to balance performance, cost, and scale as it accelerates the Gemini 3 rollout.
