On Sept. 24, 2024, Google announced the launch of Gemini 1.5 in two stable versions: Gemini 1.5 Pro and Gemini 1.5 Flash, designed to offer developers significant improvements in performance while reducing production costs. These new models bring enhanced capabilities in areas such as code generation, math and reasoning, video analysis, and factual accuracy, positioning them as a powerful tool for developers across a wide range of applications.
Key Features of Gemini 1.5 Pro and Flash
Google’s Gemini 1.5 models are intended to make it easier for developers to integrate advanced AI capabilities into their applications, with improvements over previous versions:
- Gemini 1.5 Pro offers lower costs and improved performance for developers. Google has reduced the price by over 50% compared to earlier versions, making it more accessible. In addition to the price reduction, the model now offers three times higher rate limits and lower latency than its predecessors. This can help developers scale their applications more efficiently and reduce operational costs.
- Both Gemini 1.5 Pro and Gemini 1.5 Flash show significant gains in factuality and a reduction in model hallucinations, improving the accuracy of responses. They also show improved multilingual understanding, with support for 102 languages, and stronger capabilities in SQL generation, audio understanding, and document analysis.
- One of the major upgrades is in the area of instruction following. Both models have been trained to better understand and follow user commands, making them more efficient for conversational AI applications.
- The Summarization Lengths have been reduced to optimize output, while developers building chat-based products can now access options to enhance conversational abilities within the API.
Pricing Changes and Rate Limits
To further ease adoption, Google has implemented a significant pricing change for Gemini 1.5 Pro:
- From October 1, the prices for Gemini 1.5 Pro API on prompts under 128,000 tokens will be significantly reduced:
- 64% reduction for input tokens.
- 52% reduction for output tokens.
- 64% reduction for incremental cached tokens.
Additionally, the rate limits for the paid tiers of both models have been increased, making the API more flexible and scalable for developers:
- 1.5 Flash: Increased rate limits to 2,000 requests per minute (RPM).
- 1.5 Pro: Increased rate limits to 1,000 RPM, up from 360 RPM in previous versions.
These changes offer developers the ability to process more requests at faster speeds while lowering operational costs.
Gemini 1.5 Flash-8B: Smaller Experimental Model
Google also launched an experimental version of the Gemini 1.5 Flash model called Flash-8B. This is a smaller model within the Flash family, with lower benchmark numbers but still providing significant improvements across both text-based and multimodal use cases. This version is designed for developers seeking a more lightweight option for applications that don’t require the full scale of the Pro model.
All the new Gemini 1.5 models, including Pro and Flash, are available at the Google AI Studio and can be accessed via the Gemini API.
Google vs. OpenAI: Competition Heats Up
While Google has made strides with the Gemini 1.5 release, its biggest competitor, OpenAI, is also advancing its AI capabilities. OpenAI has started rolling out a new Advanced Voice feature for ChatGPT users, which allows for more intuitive, humanlike communication with AI. As part of this update, OpenAI introduced five new voices—Arbor, Maple, SXol, Spruce, and Vale—adding to existing options like Breeze, Juniper, Cove, and Ember.
Both companies continue to enhance their offerings, but with the Gemini 1.5 launch, Google is positioning itself as a major player in the AI space for developers, with its improved models and cost reductions setting it apart in the increasingly competitive landscape.
