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
Hacker News·14d ago·5 min read
Gemini last models: temperature, top_p, and top_k are deprecated and ignored
TL;DR
Gemini 3.6 Flash and 3.5 Flash-Lite models have been released, offering improved performance on complex tasks while reducing costs and token usage. Key API changes are introduced, requiring migration for existing users.
✦ Why It Matters
Update your applications to use the latest Gemini models for improved performance and cost savings.
Key Takeaways
- Migrate to Gemini 3.6 Flash for better code generation.
- Use Gemini 3.5 Flash-Lite for high-volume data tasks.
- Remove deprecated parameters from your API requests.
- Leverage improved multimodal capabilities in your applications.
Full Summary
Gemini 3.6 Flash and 3.5 Flash-Lite are the latest models from Google, designed for production use with significant improvements over their predecessors. Gemini 3.6 Flash excels in complex agentic and multimodal tasks, reducing token usage and costs, while Gemini 3.5 Flash-Lite is optimized for speed and low-cost execution.
Both models support a 1M token context window and come with new API requirements, including the removal of deprecated parameters like temperature, top_p, and top_k. The 3.6 model enhances code generation quality and instruction following, while the 3.5 model focuses on high-volume data processing.
These updates aim to streamline workflows and improve the overall user experience in AI applications.