Google released Gemini 3 Flash worldwide on December 17, making the smaller, faster member of its Gemini 3 family the default model in the Gemini app and in the AI Mode of Google Search. The model replaces Gemini 2.5 Flash in both products and is available to developers through the Gemini API, Google AI Studio and Vertex AI.
The launch comes about a month after Google introduced Gemini 3 Pro, its flagship model, in November. Flash models are Google’s mid-tier offering, designed to run faster and cost less than the Pro line while keeping most of the capability.
Benchmarks that approach the flagship
Google says Gemini 3 Flash scores 90.4 percent on GPQA Diamond, a benchmark of graduate-level science questions that is widely used to compare reasoning ability across frontier models. According to the company, the new Flash model outperforms the previous generation’s flagship, Gemini 2.5 Pro, on most benchmarks while running roughly three times faster.
Those figures are Google’s own and have not yet been independently reproduced at the time of the launch. Benchmark results also tend to reflect specific test conditions, so they are a guide to relative capability rather than a measure of performance on any particular real-world task. Still, a mid-tier model matching or beating the prior flagship is consistent with the pattern seen across the industry in 2025, where each generation’s smaller models have closed in on the previous generation’s largest.
Pricing aimed at volume
For developers, Gemini 3 Flash is priced at $0.50 per million input tokens and $3 per million output tokens. That places it well below Gemini 3 Pro and below the top models from OpenAI and Anthropic, and it positions Flash as the model Google expects most production applications to use. Google describes the model as intended for high-volume, latency-sensitive workloads such as coding assistants, agents that call tools repeatedly, and consumer features that need fast responses.
Google also said Gemini 3 Flash supports the same context window and multimodal inputs as the rest of the Gemini 3 family, including images, audio, video and documents.
Early adopters and usage
Google named several companies that had already integrated Gemini 3 Flash before the public launch, including the developer tools maker JetBrains, the design platform Figma and the hedge fund Bridgewater Associates. The company said that since the Gemini 3 family launched, its API is processing more than one trillion tokens a day, a figure that indicates the scale of demand but does not break out usage by model or customer.
The release underlines a strategy that Google has followed throughout 2025: shipping a frontier model first, then quickly following with a cheaper version that carries most of the capability to a much larger user base. With Flash now the default for the hundreds of millions of people who use the Gemini app and AI Mode in Search, the model will likely become the version of Gemini 3 that most people actually encounter.
