https://deepmind.google/technologies/gemini/flash/ Jump to Content Google DeepMind Search [ ] Search Close Google DeepMind * About + Overview -- Find out more about Google DeepMind. + Responsibility & Safety -- We want AI to benefit the world, so we must be thoughtful about how it's built and used. + Education -- Broadening participation in science and protect the future of research and education. + Careers -- Join a community with a unique mission. + Press -- Toolkits and resources. * Research * Technologies + Overview -- Our next generation AI systems are solving some of the hardest scientific and engineering challenges of our time. + Gemini -- The Gemini family of models are the most general and capable AI models we've ever built. o Gemini models -- The Gemini family of models are the most general and capable AI models we've ever built. o Ultra -- Our largest model for highly complex tasks. o Pro -- Our best model for general performance across a wide range of tasks. o Flash -- Our lightweight model, optimized for speed and efficiency. o Nano -- Our most efficient model for on-device tasks. * Impact * Discover + Overview -- Discover our latest breakthroughs. See how we're shaping the future. Hear how AI can transform our world. + Blog -- Discover our latest AI breakthroughs, projects, and updates. + Events -- Meet our team and learn more about our research. + The Podcast -- Uncover the extraordinary ways AI is transforming our world on DeepMind: The Podcast. + Visualising AI -- Expanding our vision of what AI looks like. Search [ ] Search Close Gemini Flash Our lightweight model, optimized for speed and efficiency Build with Gemini Lightweight, fast and cost-efficient while featuring multimodal reasoning and a breakthrough long context window of up to one million tokens. Lightweight, fast and cost-efficient while featuring multimodal reasoning and a breakthrough long context window of up to one million tokens. Performance in a flash Designed to be fast and efficient to serve at scale. [r7-wJ4Lt5AOy08JKQhILmDzUpBdjUQmvufQiUrfozTrZbhrYw_kSU5] Built for speed Sub-second average first-token latency for the vast majority of developer and enterprise use cases. [aBlX8sSh7b6qwcj204MVKCFg3JJN07j8kKOph_IfX1AuxXvhp_9VXD] Quality at lower cost On most common tasks, 1.5 Flash achieves comparable quality to larger models, at a fraction of the cost. [F4z2QsbpyB7g7EXeUjtN77ijJ8woPPKFg9CFT15kOSjS7R0GNpVGmX] Long-context understanding Process hours of video and audio, and hundreds of thousands of words or lines of code. Pause Play Longer context Flash has a one-million-token context window by default, which means you can process one hour of video, 11 hours of audio, codebases with more than 30,000 lines of code, or over 700,000 words. [dXR2oGh_RGyTdBUj-APa_ElaYAZWSSmXPtJFsIwSfACQ8tBK9ruk8pltCorAXTdh-BKOrz6Sdm4ReQky93QlkfzjVQaxRGwJroGWYENkZ2gl-XUO] Build with 1.5 Flash Relentless innovation Our research team is continually exploring new ideas at the frontier of AI, building innovative products that show consistent progress on a range of benchmarks. Our newest model is Gemini 1.5 Flash. Gemini Gemini Gemini 1.5 Gemini Capability Benchmark Description 1.0 1.0 Pro 1.5 Pro Ultra Flash (Feb 2024) General MMLU Representation of questions in 57 subjects (incl. STEM, humanities, and others) Gemini Representation Gemini Gemini 1.5 Gemini of questions in 1.0 1.0 Pro 1.5 General MMLU 57 subjects Pro Ultra Flash (incl. STEM, (Feb humanities, and 71.8% 83.7% 2024) 78.9% others) 81.9% Code Natural2Code Python code generation. Held out dataset HumanEval-like, not leaked on the web Gemini Python code Gemini Gemini 1.5 Gemini generation. Held 1.0 1.0 Pro 1.5 Code Natural2Code out dataset Pro Ultra Flash HumanEval-like, (Feb not leaked on 69.6% 74.9% 2024) 77.2% the web 77.7% Math MATH Challenging math problems (incl. algebra, geometry, pre-calculus, and others) Gemini Challenging math Gemini Gemini 1.5 Gemini problems (incl. 1.0 1.0 Pro 1.5 Math MATH algebra, Pro Ultra Flash geometry, (Feb pre-calculus, 32.6% 53.2% 2024) 54.9% and others) 58.5% Reasoning GPQA (main) Challenging dataset of questions written by domain experts in biology, physics, and chemistry Challenging Gemini dataset of Gemini Gemini 1.5 Gemini questions 1.0 1.0 Pro 1.5 Reasoning GPQA (main) written by Pro Ultra Flash domain experts (Feb in biology, 27.9% 35.7% 2024) 39.5% physics, and chemistry 41.5% Reasoning Big-Bench Hard Diverse set of challenging tasks requiring multi-step reasoning Gemini Diverse set of Gemini Gemini 1.5 Gemini challenging 1.0 1.0 Pro 1.5 Big-Bench tasks requiring Pro Ultra Flash Hard multi-step (Feb reasoning 75.0% 83.6% 2024) 85.5% 84.0% Multilingual WMT23 Language translation Gemini Gemini Gemini 1.5 Gemini 1.0 1.0 Pro 1.5 Multilingual WMT23 Language Pro Ultra Flash translation (Feb 71.7 74.4 2024) 74.1 75.2 Image MMMU Multi-discipline college-level reasoning problems Gemini Gemini Gemini 1.5 Gemini Multi-discipline 1.0 1.0 Pro 1.5 Image MMMU college-level Pro Ultra Flash reasoning (Feb problems 47.9% 59.4% 2024) 56.1% 58.5% Image MathVista Multi-discipline college-level reasoning problems Gemini Gemini Gemini 1.5 Gemini Mathematical 1.0 1.0 Pro 1.5 MathVista reasoning in Pro Ultra Flash visual contexts (Feb 45.2% 53.0% 2024) 54.3% 52.1% Audio FLEURS (55 languages) Automatic speech recognition (based on word error rate, lower is better) Gemini Automatic speech Gemini Gemini 1.5 Gemini recognition 1.0 1.0 Pro 1.5 Audio FLEURS (55 (based on word Pro Ultra Flash languages) error rate, (Feb lower is better) 6.4 6.0 2024) 9.8 6.6 Video EgoSchema Video question answering Gemini Gemini Gemini 1.5 Gemini 1.0 1.0 Pro 1.5 Video EgoSchema Video question Pro Ultra Flash answering (Feb 55.7% 61.5% 2024) 63.5% 63.2% For developers Build with Gemini Integrate Gemini models into your applications with Google AI Studio and Google Cloud Vertex AI. [AqSInSAd3T7lgCxZYAn9y1oRMsrCFQj5e7QDmM2Ol9Ef8DuBLgn82dASBLsBf2otLaPo1OVaoEkjr1VN2k6CTS-U98f] Build Google AI Studio An easy way to develop model prompts and build quickly with the Gemini API. [cR-QyoMzpycbrSN5BD2ADq0uPVaUn-QFiWy3CYuxGxJazv55bR3zVE58dwI8aPzFY3eYbz-oBVRwzjU6O6E_h-irWCf] Build Vertex AI Purpose-built tools for data scientists and machine learning engineers. [gemini-new] Get the latest updates Sign up for news on the latest innovations from Google DeepMind. Email address [ ] Please enter a valid email (e.g., "name@example.com") I accept Google's Terms and Conditions and acknowledge that my information will be used in accordance with Google's Privacy Policy. Sign up Explore our other teams and product areas * # Google AI * # Google Research * # Google Cloud * # LABS.GOOGLE Footer links Follow us * * * * * About * About Google DeepMind * Responsibility & Safety * Education * Careers * Press Learn more * Research * Technologies * Impact * Discover Newsletter Email (e.g., name@example.com) [ ] Please enter a valid email (e.g., "name@example.com") I accept Google's Terms and Conditions and acknowledge that my information will be used in accordance with Google's Privacy Policy. Sign up * About Google * Google products * Privacy * Terms * Cookies management controls