A Major Leap in Accessible AI Intelligence
Just one month after launching its flagship Gemini 3 Pro model, Google has fired the next salvo in the intensifying AI race. The company has officially launched Gemini 3 Flash and is making it the new default model for hundreds of millions of users worldwide. This strategic move brings the advanced "frontier intelligence" of the Gemini 3 family to the masses, combining Pro-level reasoning with the speed and efficiency the Flash series is known for.
Starting today, Gemini 3 Flash is rolling out as the default model in the Gemini app globally, replacing Gemini 2.5 Flash at no cost to users. Simultaneously, it is becoming the default model powering AI Mode in Google Search worldwide, promising smarter and faster answers to complex queries. For developers and enterprises, the model is available immediately via API and platforms like Vertex AI, with companies like Salesforce, Figma, and JetBrains already deploying it.
Why Gemini 3 Flash is a Game Changer
Google positions Gemini 3 Flash as "frontier intelligence built for speed at a fraction of the cost". It's not just an incremental update; it represents a significant shift in the performance-to-speed paradigm. The model is designed to handle the reasoning demands of complex tasks while delivering responses with the low latency expected from a "Flash" model.
Unmatched Speed
Benchmarks show it is 3x faster than Gemini 2.5 Pro while outperforming it in quality. It's optimized for high-frequency, real-time interactions.
Advanced Multimodality
Can analyze and reason over video, audio, images, and text in near real-time. Upload a short video for swing tips or a sketch for instant analysis.
Elite Agentic Coding
Achieves a 78% score on SWE-bench Verified, outperforming not only its predecessors but also Gemini 3 Pro in this coding agent benchmark.
Benchmark Dominance
Gemini 3 Flash doesn't just beat its predecessors; it competes with and often surpasses the current best models from competitors, including OpenAI's recently rushed-out GPT-5.2.
- MMMU Pro (Multimodal Understanding): 81.2% - outperforms GPT-5.2 (79.5%).
- Humanity's Last Exam (Academic Reasoning): 33.7% without tools - nearly matches GPT-5.2 (34.5%).
- GPQA Diamond (Scientific Knowledge): 90.4% - rivals much larger frontier models.
- It also "significantly outperforms" the previous best-in-class Gemini 2.5 Pro across multiple benchmarks.
Global Rollout & How to Access It
Global, Immediate
Global Rollout Started
Available Now in Preview
For Consumers (Free): If you use the Gemini app, you should see the upgrade automatically. It replaces Gemini 2.5 Flash as the default. In the app, you'll find two modes: "Fast" (Gemini 3 Flash for quick answers) and "Thinking" (for complex problems). For search, when you use AI Mode on google.com, it is now powered by Gemini 3 Flash globally, providing more nuanced and visually organized answers.
Expanded Pro Access in the U.S.: Alongside the Flash rollout, Google is expanding access to its more powerful models in Search for U.S. users. Everyone in the U.S. can now select "Thinking with 3 Pro" in the AI Mode menu for complex tasks, and more users gain access to the "Nano Banana Pro" image generation model.
For Developers & Businesses: The model is available in preview now through the Gemini API in Google AI Studio, Google Antigravity, Gemini CLI, Vertex AI, and Gemini Enterprise. Pricing is set at $0.50 per 1 million input tokens and $3.00 per 1 million output tokens. Google highlights that while the per-token cost is slightly higher than its predecessor, its efficiency (using 30% fewer tokens on average) and vastly superior performance create a better value proposition.
