Today, Google introduced Gemini 3 Flash, offering top-tier intelligence at lightning speed and an incredibly low cost.

Last month, Google launched Gemini 3, featuring both Gemini 3 Pro and Gemini 3 Deep Think modes.

Gemini 3 Flash combines the professional reasoning capabilities of Gemini 3 with Flash-level latency, efficiency, and cost benefits, marking it as Google's most remarkable model for agent workflows to date.

Beginning today, Gemini 3 Flash will be accessible to users:

For developers leveraging the Gemini API in Google AI Studio, Gemini CLI, and the innovative agent development platform Google Antigravity.

Available to all users through the Gemini app and AI mode in Search.

For enterprises utilizing Vertex AI and Gemini Enterprise.

Gemini 3 Flash showcases that speed and scalability need not compromise intelligence. It exhibits cutting-edge performance in doctoral-level reasoning and knowledge benchmarks, such as GPQA Diamond (90.4%) and Humanity's Last Exam (33.7%), matching the capabilities of larger advanced models and significantly outperforming the current top 2.5 version model, Gemini 2.5 Pro, across various benchmarks.

Moreover, it achieved an 81.2% score on the MMMU Pro test, comparable to the performance of Gemini 3 Pro.

Beyond its advanced reasoning and multimodal processing capabilities, Gemini 3 Flash is engineered for maximum efficiency, pushing the boundaries of the Pareto frontier between quality, cost, and speed.

When operating at its highest thinking levels, Gemini 3 Flash can dynamically adjust its thinking time. For more intricate scenarios, it may require extended thinking periods, but based on typical traffic tests, it uses an average of 30% fewer tokens than 2.5 Pro, executing daily tasks more accurately and with enhanced performance.

Gemini 3 Flash outperforms 2.5 Pro with 3x the speed (based on Artificial Analysis benchmarks) at a substantially lower cost. Gemini 3 Flash is priced at $0.50 per million input tokens and $3 per million output tokens.

Designed for iterative development, Gemini 3 Flash features the professional coding performance of Gemini 3 with exceptionally low latency.

In the SWE-bench Verified benchmark, which assesses coding agent capabilities, Gemini 3 Flash achieved an impressive 78%, outperforming not only the 2.5 series but also Gemini 3 Pro.

Gemini 3 Flash's robust performance in reasoning, tool utilization, and multimodal functions makes it perfect for developers seeking more sophisticated video analysis, data extraction, and visual question answering, enabling smarter applications—such as gaming assistants or A/B testing experiments—that require both rapid responses and in-depth reasoning.

Companies like JetBrains, Bridgewater Associates, and Figma have already begun implementing it to revolutionize their operations. Denis Shiryaev, Chief AI Ecosystem Tool Developer at JetBrains, noted that in controlled production environments, Gemini 3 Flash consistently adheres to user token budgets, ensuring complex multi-step AI Agents perform swiftly, predictably, and at scale.

Building on the reasoning capabilities of Gemini 3 Pro, Gemini 3 Flash's AI mode can more effectively decipher the subtleties of questions and seamlessly integrate research with immediate action.

Gemini 3 Flash is now available for preview through the Gemini API in Google AI Studio, Google Antigravity, Vertex AI, and Gemini Enterprise. Users can also access it via other developer tools, such as Gemini CLI and Android Studio.

References:

https://blog.google/products/gemini/gemini-3-flash/