We’re launching Lyria 3.5 in Google Flow Music, with advances across musicality, lyrics, vocals, and creative control
deepmind.google·6d ago
AI Engineering
DeepMind Blog·Jun 17·5 min read
We’re expanding our Gemini 2.5 family of models
TL;DR
Developers needed efficient AI models that balance performance with cost and speed. DeepMind has released Gemini 2.5 Flash and Pro as stable versions, along with a new model called 2.5 Flash-Lite, which is designed to be the fastest and most cost-effective.
✦ Why It Matters
Engineers can leverage the new Gemini 2.5 models to enhance application performance while managing costs effectively.
Key Takeaways
- Explore the new Gemini 2.5 Flash and Pro models for production use.
- Utilize Gemini 2.5 Flash-Lite for cost-efficient, high-speed tasks.
- Leverage the 1 million-token context length for complex applications.
- Provide feedback on the preview version to influence future updates.
Full Summary
The Gemini 2.5 family of models has been designed as hybrid reasoning models that balance cost and speed effectively. The newly released 2.5 Flash and Pro models are now stable and available for developers to use in production, with organizations like Snap already implementing them.
Additionally, the preview of Gemini 2.5 Flash-Lite offers the fastest and most cost-efficient option yet, outperforming its predecessor, 2.0 Flash-Lite, in various benchmarks. It excels in high-volume tasks such as translation and classification, featuring lower latency and a 1 million-token context length.
Users can access these models through Google AI Studio and Vertex AI, encouraging feedback for further improvements.