Technology media outlet The Decoder published a blog post on Sept. 24, reporting thatGoogleUnder the Upgrade banner Gemini 1.5 AI model, launched Gemini-1.5-Pro-002 and Gemini-1.5-Flash-002It is less expensive, more powerful, and more responsive than previous versions.
Lower cost
Google lowered token input and output fees by up to 50% for Gemini-1.5-Pro-002 and Gemini-1.5-Flash-002, increased rate limits for both models, and reduced latency.
New pricing effective October 1, 2024
Better performance
Quoting from the press release, the performance of the new model is attached below:
- In the more challenging MMLU-Pro benchmark, the model's performance improved by about 7%.
- Math performance was significantly improved by 20% in the MATH and HiddenMath benchmarks.
- Visual and code-related tasks also improved, by 2-7% in the visual understanding and Python code generation assessments.
Google claims that the models can now provide more helpful responses while maintaining content security standards. The company has improved the output style of the models based on feedback from developers, aiming for more accurate and cost-effective use.
Other improvements
Google has also upgraded its Gemini 1.5 experimental model, released in August, with the introduction of the Gemini-1.5-Flash-8B-Exp-0924 Upgraded version with further enhancements to text and multimodal applications.
Users can access the new Gemini models through Google AI Studio, the Gemini API, and Vertex AI (for Google Cloud customers). A chat-optimized version of Gemini 1.5 Pro-002 for Gemini Advanced users is coming soon.