Gemini 3.1 フラッシュライト — Google DeepMind

Gemini 3.1 Flash-Lite — Google DeepMind Skip to main content Gemini 3.1 Flash-Lite Best for high-volume tasks that need efficiency and intelligence Build with Gemini Your browser does not support the video tag. Your browser does not support the video tag. Introducing 3.1 Flash-Lite, a scalable thinking model for high-volume tasks at low cost and latency. Capabilities Hands-on Showcase Performance Model information Try Gemini 3.1 Flash-Lite Handles tasks of varying complexity like coding, UI generation and translation with high quality Slide 1 of 4 Flexible reasoning levels Delivers improved reasoning and output quality, allowing users to select the level of thinking they want to use. Low latency Tackles high-volume tasks with faster response times. Tool use Delivers high throughput with quality using search grounding and enhanced instruction following. Cost-efficient Our most cost efficient model yet in the 3 series. Hands-on Explore what you can do with Gemini 3.1 Flash-Lite Slide 1 of 2 Your browser does not support the video tag. Your browser does not support the video tag. Develop real-time e-commerce categories Gemini 3.1 Flash-Lite populates an e-commerce wireframe with hundreds of products across multiple categories in seconds. Your browser does not support the video tag. Your browser does not support the video tag. Build a generative weather dashboard Gemini 3.1 Flash-Lite builds live weather dashboards on demand, pulling real-time forecasts and historical data into dynamic visualizations. Showcase Slide 1 of 4 Google’s model has demonstrated unparalleled instruction-following capabilities and speed in its class, achieving 20% higher success rate and 60% faster inference times than our previous model. It's enabling Latitude to deliver sophisticated storytelling to a much wider audience than would have otherwise been possible. Kolby Nottingham Head of AI, Latitude 3.1 Flash-Lite is a remarkably competent model. It is lightning fast, but still somehow finds a way to follow all instructions. It is great at tool calling and can rapidly explore codebases in a fraction of the time of bigger models. We have a wide variety of multimodal labeling use cases, at dramatic scale, we’ve found Flash-Lite to be an unlock for our ability to bring insight to more data at even larger scale. The intelligence to speed ratio is unparalleled in any other model. Andrew Carr Chief Scientist, Cartwheel By integrating 3.1 Flash-Lite into our classification pipeline, Whering has achieved 100% consistency in item tagging, providing a highly reliable foundation for our label assignment. 3.1 Flash-Lite’s ability to deliver certain, repeatable results, even on complex fashion categories, has streamlined our data labelling process and increased our confidence in the structured outputs. Bianca Rangecroft CEO, Whering As a root orchestration and content engine 3.1 Flash-Lite consistently delivered sub-10 second completions with near-instant streaming, ~97% structured output compliance, and 94% intent routing accuracy. For high-throughput AI products, it offers an exceptional balance of speed, instruction fidelity, and cost efficiency. Kaan Ortabas Co-Founder, HubX Performance 3.1 Flash-Lite performs significantly better than 2.5 Flash across a number of key benchmarks, including general quality, reasoning, translation and factuality. Benchmark Notes Gemini 3.1 Flash-Lite High Gemini 2.5 Flash Dynamic Gemini 2.5 Flash-Lite Dynamic GPT-5 mini High Claude 4.5 Haiku Extended Thinking Grok 4.1 Fast Reasoning Input price $/1M tokens, no caching Lower is better $0.25 $0.30 $0.10 $0.25 $1.00 $0.20 Output price $/1M tokens Lower is better $1.50 $2.50 $0.40 $2.00 $5.00 $0.50 Output speed Tokens / s 363 249 366 71 108 145 Humanity’s Last Exam Academic reasoning (full set, text + MM) No tools 16.0% 11.0% 6.9% 16.7% 9.7% 17.6% GPQA Diamond Scientific knowledge No tools 86.9% 82.8% 66.7% 82.3% 73.0% 84.3% MMMU-Pro Multimodal understanding and reasoning No tools 76.8% 66.7% 51.0% 74.1% 58.0% 63.0% CharXiv Reasoning Information synthesis from complex charts 73.2% 63.7% 55.5% 75.5% (+ python) 61.7% 31.6% Video-MMMU Knowledge acquisition from videos 84.8% 79.2% 60.7% 82.5% — 74.6% SimpleQA Verified Parametric knowledge 43.3% 28.1% 11.5% 9.5% 5.5% 19.5% FACTS Benchmark Suite Factuality benchmark across grounding, parametric, search, and MM. 40.6% 50.4% 17.9% 33.7% 18.6% 42.1% MMMLU Multilingual Q&A 88.9% 86.6% 84.5% 84.9% 83.0% 86.8% LiveCodeBench Code generation (UI: 1/1/2025-5/1/2025) 72.0% 62.6% 34.3% 80.4% 53.2% 76.5% MRCR v2 (8-needle) Long context performance 128k (average) 60.1% 54.3% 30.6% 52.5% 35.3% 54.6% 1M (pointwise) 12.3% 21.0% 5.4% Not supported Not supported 6.1% Methodology: deepmind.google/models/evals-methodology/gemini-3-1-flash-lite Model information Name 3.1 Flash-Lite Status Preview Input Text Image Video Audio PDF text_snippet image videocam mic drive_pdf Output Text text_snippet image videocam mic drive_pdf Input tokens 1M Output tokens 64k Knowledge cutoff January 2025 Tool use Function calling Structured output Search as a tool Code execution Best for High-volume, latency-sensitive reasoning tasks Availability Google AI Studio Gemini API Vertex AI Documentation View developer docs Model card View model card