Google Launches Gemini 3 Flash, Promising Enhanced AI Capabilities

Google has launched its new AI model, Gemini 3 Flash, which is now the default model in the Gemini app and AI mode in search. This model is built for faster output at a lower cost and promises significant improvements over its predecessor, Gemini 2.5 Flash.

According to TechCrunch, Gemini 3 Flash outperforms its predecessor on various benchmarks, achieving a score of 33.7% on the Humanitys Last Exam benchmark without tool use, while the Gemini 3 Pro scored 37.5%.

Additionally, it scored 81.2% on the multimodality and reasoning benchmark MMMU-Pro, surpassing all competitors. The model is available globally and is designed for tasks like video analysis, data extraction, and visual Q&A.

For developers, Gemini 3 Flash can be accessed through Vertex AI and the Gemini API. Its performance on coding tasks is notable, with a score of 78% on the SWE-bench verified coding benchmark, only slightly behind GPT-5.2.

Despite a slight increase in pricing per token compared to Gemini 2.5 Flash, Google claims the new model is more efficient, processing workloads three times faster. CNET reports that Gemini 3 Flash is positioned as a smart and fast model, capable of handling everyday tasks and customer support.

This launch comes amid heightened competition with OpenAI, which has recently released its own updates. According to Ars Technica, Gemini 3 Flash has gained almost 20 points over its predecessor in coding benchmarks and is less likely to make mistakes on general knowledge questions.

Google emphasizes that Gemini 3 Flash offers a blend of intelligence and efficiency, making it a significant player in the AI landscape.