Google Quietly Releases Gemini 3.8 Flash for Coding and AI Agents

Google has quietly released the Gemini 3.8 Flash model on its DeepMind site, without a big announcement. It's based on Gemini 3.7 Flash and focuses on improving software engineering and AI agent workflows. You can adjust the effort level to balance quality, cost, and speed.

This model has a context window of up to 1 million tokens and can output up to 64K tokens. It's made for individuals, developers, and businesses, and is good for running AI agents at a low cost. Typical uses include coding, agent tasks, and handling complex knowledge. Google says it performs better in coding, handling images and videos, long documents, using computers, and science reasoning.

The model's knowledge goes up to March 2026, but some info might be older. Google warns it can sometimes make things up, and they're working on making it safer. If you use it with high reasoning, it might be slow or use more tokens.