Google releases new AI model Gemini 1.5 Flash, a fast multimodal model

2024-05-15

Google announced the launch of Gemini 1.5 Flash, a small multimodal model designed for scalable construction and processing of narrow high-frequency tasks. It features a context window of one million tokens and is now available for public preview through the Gemini API in Google AI Studio. At the same time, Gemini 1.5 Pro, which made its debut in February, has received a significant update, expanding its context window from one million tokens to two million tokens. Developers interested in this update are required to register and wait. There are significant differences between Gemini 1.5 Flash and Gemini 1.5 Pro. The former focuses on output speed and is suitable for fast tasks with low latency requirements, while the latter is more powerful, comparable to Google's large-scale 1.0 Ultra model, and is more suitable for "more general or complex, often multi-step reasoning tasks." Now, developers can choose the AI model that best suits their needs, no longer limited to a single universal approach. Different applications require different data and AI capabilities, and having multiple variants allows users to experience a more diverse range of AI-driven services. Google brings state-of-the-art AI models to developers through this approach while ensuring their performance is optimized. However, it is worth noting that Gemini 1.5 Flash may not have been trained on a large enough dataset. For developers with higher demands, upgrading to Gemini 1.5 Pro is recommended. Google's model covers the entire spectrum from lightweight Gemma and Gemma 2 to more powerful Gemini Nano, Gemini 1.5 Flash, Gemini 1.5 Pro, and Gemini 1.0 Ultra. Josh Woodward, Vice President of Google Research, pointed out, "Developers can flexibly choose between models of different sizes based on specific use cases. These models have the same multimodal input capability, long-context processing capability, and perform well in the same backend." This new small language model was unveiled within 24 hours after Google's AI competitor OpenAI released GPT-4o, a multimodal LLM for all users, including desktop applications. Currently, both Gemini 1.5 models are available for public preview in over 200 countries and regions worldwide, covering the European Economic Area, the United Kingdom, Switzerland, and more. By releasing Gemini 1.5 Flash and upgrading Gemini 1.5 Pro, Google once again demonstrates its leadership and innovation in the field of AI. With the widespread application of these advanced models, we have reason to believe that AI technology will continue to bring more convenience and progress to human society.