Google announced the release of two new versions of its popular Gemini AI models that promise a 50% price drop for developers as well as two times faster output and three times lower latency over the previous generation. Gemini-1.5-Pro-002 and Gemini-1.5-Flash-002 will initially be available to developers but will also come to the Gemini App and website.
Among the most notable features, the faster output makes these models ideal for high-demand AI tasks such as code, text, and multimodal processing. Performance has been enhanced across multiple benchmarks, showing a 20% improvement in math-related tasks and substantial gains in visual understanding and code generation.
Considering developer feedback, Google reduced the output length by 5-20%, resulting in faster responses without compromising quality. For longer, more verbose responses, developers can apply specific prompting strategies.
The previous version of Gemini Flash 1.5 beat Gemini Pro 1.5 in our testing, so it will be interesting to see if this upgrade makes a difference.
Designed for performance, flexibility, and safety
The new Gemini-1.5-Pro model is well-suited for tasks like summarizing information from large documents and understanding long videos. For developers, accessing these models via Google AI Studio and Vertex AI has been streamlined, and the improved rate limits make it easier to build with Gemini’s capabilities at scale.
In an effort to give developers more of what they want, Google has experimented with Gemini-1.5-Flash to bring even greater improvements across text and multimodal tasks. It is clear that the company is committed to evolving the AI landscape while taking to heart developers' needs. Alongside these model upgrades, Google updated privacy filters, allowing developers more flexibility in how they configure the models while maintaining content safety.
With these enhancements, Google has made the Gemini models not only more affordable but also more powerful, responsive and safer, solidifying their place as competitive tools in the AI space. These improvements continue to open new possibilities for AI applications, and Google is encouraging developers to explore the models' potential across various use cases.
In addition to the aforementioned updates, Google announced that Gemini Advanced users can expect to access a chat optimized version of Gemini 1.5 Pro-002 soon, signaling there is more to come to excite users and industry leaders alike.
The updated version of Gemini Pro 1.5 will be available to Gemini Advanced subscribers.
