Revised Gemini Models for Production, Lowered 1.5 Pro Prices, Enhanced Rate Limits, and Additional Updates – Google Developers Blog

Today, we’re releasing two updated production-ready Gemini models: Gemini-1.5-Pro-002 and Gemini-1.5-Flash-002 along with:
>50% reduced price on 1.5 Pro (both input and output for prompts <128K) 2x higher rate limits on 1.5 Flash and ~3x higher on 1.5 Pro 2x faster output and 3x lower latency Updated default filter settings These new models build on our latest experimental releases and include meaningful improvements to the Gemini 1.5 models introduced at Google I/O in May. Developers can access our latest models for free through Google AI Studio and the Gemini API. For larger organizations and Google Cloud customers, the models are also available on Vertex AI. Improved overall quality, with larger gains in math, long context, and vision. The Gemini 1.5 series features models designed for general performance across a wide range of text, code, and multimodal tasks. For example, Gemini models can synthesize information from 1000-page PDFs, answer questions about repositories containing more than 10,000 lines of code, and extract useful content from hour-long videos, among other capabilities. With the latest updates, 1.5 Pro and Flash are now better, faster, and more cost-efficient for production use. We see a ~7% increase in MMLU-Pro, a more challenging version of the popular MMLU benchmark. On MATH and HiddenMath (an internal holdout set of competition math problems) benchmarks, both models have improved by approximately ~20%. For vision and code use cases, both models also demonstrate better performance (ranging from ~2-7%) across evaluations measuring visual understanding and Python code generation. We have also enhanced the overall helpfulness of model responses while continuing to uphold our content safety policies and standards. This results in fewer refusals and more informative responses across various topics. Both models now feature a more concise style based on developer feedback, aimed at improving usability and reducing costs. For use cases like summarization, question answering, and extraction, the default output length of the updated models is ~5-20% shorter than previous iterations. For chat-based products where longer responses may be preferred, you can consult our prompting strategies guide to learn how to make the models more verbose and conversational. For more details on migrating to the latest versions of Gemini 1.5 Pro and 1.5 Flash, please visit the Gemini API models page. Gemini 1.5 Pro We continue to be impressed by the innovative and practical applications of Gemini 1.5 Pro’s 2 million token-long context window and multimodal capabilities. From video understanding to processing 1000-page PDFs, there are countless new use cases waiting to be explored. Starting October 1st, 2024, we are announcing a 64% price reduction on input tokens, a 52% price reduction on output tokens, and a 64% price reduction on incremental cached tokens for our flagship 1.5 series model, Gemini 1.5 Pro, on prompts of less than 128K tokens. Together with context caching, this drives down the cost of building with Gemini. Increased rate limits To facilitate development with Gemini, we are raising the paid tier rate limits for 1.5 Flash to 2,000 RPM and for 1.5 Pro to 1,000 RPM, up from 1,000 and 360, respectively. In the coming weeks, we anticipate continuing to increase the Gemini API rate limits so developers can fully leverage Gemini. 2x faster output and 3x less latency Alongside essential improvements to our latest models, we have recently reduced latency with 1.5 Flash and significantly increased the output tokens per second, enabling new use cases with our most powerful models. Updated filter settings Since the initial launch of Gemini in December of 2023, building a safe and reliable model has been a key focus. The latest versions of Gemini (-002 models) feature improvements in the model's ability to follow user instructions while balancing safety. We will continue to provide a suite of safety filters that developers may apply to Google’s models. For the models released today, the filters will not be applied by default, allowing developers to choose the configuration best suited to their use case. Gemini 1.5 Flash-8B Experimental updates We are releasing an enhanced version of the Gemini 1.5 model introduced in August, called “Gemini-1.5-Flash-8B-Exp-0924.” This version boasts significant performance increases across both text and multimodal use cases. It is available now through Google AI Studio and the Gemini API. We have received overwhelmingly positive feedback from developers regarding 1.5 Flash-8B, and we will continue to refine our experimental-to-production release pipeline based on this input. We are excited about these updates and look forward to seeing the innovative solutions you will create with the new Gemini models! Additionally, Gemini Advanced users will soon have access to a chat-optimized version of Gemini 1.5 Pro-002.