Gemini 3 Flash Brings Frontier Intelligence to the Fast Lane

Google just dropped Gemini 3 Flash—and it's built for speed without sacrificing smarts.

The new model delivers "frontier performance" on PhD-level reasoning benchmarks, according to the DeepMind blog. That includes 90.4% on GPQA Diamond (a benchmark testing graduate-level science questions) and 33.7% on Humanity's Last Exam. Those numbers rival much larger frontier models, Google says.

The pitch: you get Gemini 3's reasoning capabilities at Flash-level speed and cost. It's designed for coding, complex analysis, and interactive apps where latency matters.

"Gemini 3 Flash demonstrates that speed and scale don't have to come at the cost of intelligence," according to Google's announcement.

The model is now the default in the Gemini app and AI Mode in Search. Developers can access it via Google AI Studio, Gemini CLI, Vertex AI, and the new Google Antigravity platform. Since launch day of Gemini 3, Google says it's been processing over 1 trillion tokens per day on its API.