Google Releases Gemini 3 Flash, Outperforms 3 Pro Model in Speed and Coding Performance

Highlights

• Google says Gemini 3 Flash outperforms the entire 2.5 series
• It is 3X faster than Gemini 2.5 Pro while using 30 percent fewer tokens
• Enterprises can access the model via Vertex AI and Gemini Enterprise

Google released Gemini 3 Flash on Wednesday as the latest entrant in the Gemini 3 series. The artificial intelligence (AI) model joins Gemini 3 Pro and Gemini 3 Deep Think, and brings speed, efficiency, and lower token cost for users and developers. Arriving a month after the release of the previous two models, Google says the Flash variant is powerful enough to outperform 3 Pro in coding-related tasks. It is also said to be more performant compared to the entire Gemini 2.5 series.

Gemini 3 Flash: Details and Availability

In a blog post, the Mountain View-based tech giant announced the release of Gemini 3 Flash and detailed its capabilities. It is also rolling out globally to everyone via the Gemini (app and website) and AI Mode in Search. Developers will be able to access it via the Gemini application programming interface (API) in Google AI Studio, Gemini CLI, and the agentic development platform Antigravity. Businesses can find it in Vertex AI and via Gemini Enterprise.

Google's AI models with the “Flash” moniker are known for low latency and affordability, and the Gemini 3 Flash continues the tradition. However, the company has also made it more capable. It is said to be more intelligent than the entire 2.5 series, including the Gemini 2.5 Pro. In some areas, it also matches and outperforms Gemini 3 Pro.

Based on internal evaluations, the tech giant claims that the Gemini 3 Flash scored 90.4 percent on the GPQA Diamond benchmark (reasoning and knowledge), and 33.7 percent without tools on Humanity's Last Exam (academic-grade reasoning). It is also said to achieve 81.2 percent on the MMMU Pro benchmark and 78 percent on the SWE-bench Verified.

Google also said that its latest AI model can think for longer when a user presents it with a more complex query; however, it uses 30 percent fewer tokens on average when compared to the 2.5 Pro. The tech giant says this not only makes the model more efficient, but it also impacts the cost-effectiveness.

Coming to pricing, Gemini 3 Flash costs $0.50 (roughly Rs. 45) per million input tokens and $3 (roughly Rs. 271) per million output tokens. The audio input remains locked at $1 (roughly Rs. 90.5) per million tokens. Interestingly, while the pricing makes it significantly cheaper than Gemini 3 Pro, the 3 Flash model is slightly more expensive than the 2.5 Pro variant.