EmbeddingGemma 2: an open, lightweight multimodal embedding model
deepmind.google·6d ago
AI Engineering
DeepMind Blog·Jul 21·5 min read
Introducing Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber
TL;DR
DeepMind has launched Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber, enhancing AI agent performance with improved token efficiency and lower costs. These models cater to diverse needs, from coding to cybersecurity applications.
✦ Why It Matters
Evaluate and integrate the new Gemini models to enhance your AI applications today.
Key Takeaways
- Utilize Gemini 3.6 Flash for cost-effective coding tasks.
- Implement 3.5 Flash-Lite for high-speed document processing.
- Leverage 3.5 Flash Cyber for enhanced cybersecurity measures.
- Explore the Gemini API for seamless integration into existing workflows.
Full Summary
Gemini 3.6 Flash builds on user feedback from 3.5 Flash, achieving a 17% reduction in output token usage while enhancing coding and knowledge work capabilities. It operates at a lower cost of $1.50 per million input tokens and $7.50 per million output tokens, making it more economical for developers.
The 3.5 Flash-Lite model is optimized for speed, delivering 350 output tokens per second, and is priced at $0.30 per million input tokens. Additionally, 3.5 Flash Cyber is tailored for cybersecurity, efficiently identifying and fixing vulnerabilities through the CodeMender platform.
These advancements not only improve performance metrics but also incorporate enhanced safety features to mitigate misuse. Overall, the new models aim to streamline agentic workflows and bolster security measures in software development.