On December 17, 2025, Google announced the general availability of Gemini 3 Flash, a high-speed AI model that marks a significant shift in the industry by offering "Pro-level" reasoning at "Flash" speeds. Replacing the previous Gemini 2.5 Flash as the default engine for the Gemini app and AI Search, the new model effectively bridges the gap between massive, slow frontier models and lightweight, fast ones.

The most striking achievement of Gemini 3 Flash is its performance against its own predecessors. Google confirmed that this lightweight model actually outperforms Gemini 2.5 Pro—last year’s flagship—across 18 of 20 major benchmarks. Most notably, it achieved a score of 90.4% on the GPQA Diamond benchmark, a test specifically designed to measure PhD-level scientific reasoning, a feat previously reserved only for the largest and most expensive AI models.

Beyond raw intelligence, the model is engineered for extreme efficiency. Gemini 3 Flash is clocked at three times the speed of Gemini 2.5 Pro while utilizing roughly 30% fewer tokens to solve the same complex tasks. This architectural breakthrough allows for near-instantaneous responses even during intense reasoning cycles, making it ideal for high-frequency workflows like real-time customer support or live coding assistance.

For developers and enterprises, the pricing structure is a major highlight. At $0.50 per 1 million input tokens and $3.00 per 1 million output tokens, the model costs less than a quarter of the price of Gemini 3 Pro. Google’s aggressive pricing strategy is clearly aimed at high-volume agentic applications, where cost-to-performance ratios often determine the viability of a product.

A new feature accompanying the launch is Configurable Reasoning Levels. Users and developers can now toggle between four thinking modes: Minimal, Low, Medium, and High. This allows for a "Fast" experience for simple chat queries and a "Thinking" mode for deep problem-solving. An "Auto" mode is also available, where the model intuitively decides how much "thought" to put into a response based on the complexity of the prompt.

The model's multimodal capabilities have also seen a massive upgrade. Gemini 3 Flash supports a 1 million token context window, capable of processing hours of video, thousands of lines of code, or massive PDF libraries natively. It is specifically optimized for Google’s new Antigravity agentic coding platform, where it achieved a state-of-the-art 78% score on SWE-bench Verified, actually surpassing the more expensive Gemini 3 Pro in specific coding tasks.

As of today, Gemini 3 Flash is live for all consumers in the Gemini app and is available in preview via the Gemini API, Google AI Studio, and Vertex AI. With this launch, Google has set a new industry benchmark: the "Flash" tier is no longer just for basic tasks—it is now a frontier-class competitor that challenges the necessity of slower, more expensive flagship models.