Google has officially launched Gemini 3 Flash, a new iteration of its artificial intelligence model designed to deliver significant improvements in both intelligence and operational efficiency. This release follows the recent introduction of the Gemini 3 Pro model, marking a swift advancement in Google’s AI development pipeline. The Gemini 3 Flash model is set to be integrated across various Google products, including the Gemini app and search engine, while also becoming immediately available to developers through its API and cloud platforms.

Early benchmarks indicate that Gemini 3 Flash surpasses its predecessor, Gemini 2.5 Flash, in key areas of reasoning and academic understanding. The model demonstrated notable gains in tests such as GPQA Diamond and MMMU Pro, even outperforming the Gemini 3 Pro model in some instances. A particularly striking improvement was observed in the Humanity’s Last Exam (HLE) benchmark, where Gemini 3 Flash achieved a score of 33.7 percent without utilizing external tools, tripling the performance of previous models and nearing the capabilities of Gemini 3 Pro.

Google is also highlighting Gemini 3 Flash’s enhanced coding abilities, with benchmarks supporting these claims. While Google has historically positioned its Pro models as superior for code generation, Gemini 3 Flash has significantly closed the gap. In the widely recognized SWE-Bench Verified test, the new model showed an improvement of nearly 20 points compared to the Gemini 2.5 branch, underscoring its growing proficiency in this domain.

The Gemini 3 Flash model also addresses a critical aspect of AI performance: accuracy in general knowledge retrieval. In the Simple QA Verified test, Gemini 3 Flash achieved a score of 68.7 percent, a substantial leap from the 28.1 percent recorded by the last Flash model, bringing it much closer to the accuracy levels of Gemini 3 Pro.

Beyond raw intelligence, Gemini 3 Flash emphasizes efficiency and cost-effectiveness for developers. While offering performance improvements over Gemini 2.5 Pro, it operates up to three times faster. This enhanced speed translates to more economical usage for developers, with input tokens priced at $0.50 per million and output tokens at $3 per million. Although this represents an increase from the previous Gemini 2.5 Flash pricing, it remains considerably more affordable than the Pro models.

The underlying multimodal capabilities that characterized Gemini 3 Pro, such as the generation of interactive simulations and diverse content formats, are reportedly retained in Gemini 3 Flash. This means users and developers can expect a blend of speed, intelligence, and advanced creative potential from this latest offering, further solidifying Google’s commitment to pushing the boundaries of AI accessibility and application.

Read Original

Keywords: Gemini 3 Flash, Google AI, AI model, artificial intelligence, LLM, AI efficiency, AI performance, AI development