Google has launched Gemini 3 Flash, an upgraded model of its AI system, which follows the recent introduction of the Pro version. This release enhances performance across the Gemini app and search, while also becoming immediately available through the Gemini API and other developer tools. The new model claims to be faster and more capable than the previous 2.5 Flash, showing notable improvements in various benchmarks. For instance, in reasoning and academic tests, Gemini 3 Flash outperforms 2.5 Flash, particularly in Humanity’s Last Exam, where it achieved a score of 33.7 percent—three times better than its predecessor. Gemini 3 Flash also shows significant advancements in coding capabilities. It has scored nearly 20 points higher than 2.5 Flash in the SWE-Bench Verified test, suggesting it can now compete closely with the Pro models that Google has touted for coding tasks. In the Simple QA Verified test, it scored 68.7 percent, a considerable leap from the 28.1 percent of the last Flash model. While the Flash model performs nearly on par with the Gemini 3 Pro in many assessments, it is also positioned as a more efficient option. The efficiency of Gemini 3 Flash is underscored by its cost-effectiveness. Developers will pay $0.50 for one million input tokens and $3 for output tokens, marking an increase from the previous model's prices but still lower than the Pro model's rates of $2 and $12 respectively. This makes Gemini 3 Flash an attractive option for developers looking to leverage its capabilities without incurring high costs. The model retains advanced features like interactive simulations and multimodal content generation, further broadening its utility.