What happened
In May 2026, on May 19, during Google I/O 2026, Google launched Gemini 3.5 Flash, the first model in the new Gemini 3.5 family. The company presented it as its strongest model yet for coding and autonomous AI agents, combining "frontier intelligence with action."
Unlike earlier preview launches, 3.5 Flash arrived generally available (GA) the same day, for everyday users and developers alike.
Where it is available
- For everyone: it became the default model in the Gemini app (including free users) and in AI Mode in Google Search, worldwide.
- For developers: in the agent-first development platform Google Antigravity, the Gemini API (Google AI Studio) and Android Studio.
- For enterprises: in Gemini Enterprise Agent Platform and Gemini Enterprise.
- The same day, GitHub announced the model as generally available in GitHub Copilot.
Performance
According to Google, 3.5 Flash beats Gemini 3.1 Pro, until then the company's frontier model, on nearly all coding and agent benchmarks. Some published numbers:
- Terminal-Bench 2.1: 76.2%;
- GDPval-AA: 1656 Elo;
- MCP Atlas: 83.6%;
- CharXiv Reasoning (multimodal understanding): 84.2%.
Google also says that, in output tokens per second, it is about 4 times faster than other frontier models. That speed is central for work where many agents run at once. On stage, Google showed agents splitting tasks to build an entire operating system inside Antigravity.
Technical details for migration
The official developer guide includes practical points:
- a 1 million token context window and up to 65k output tokens;
- the default thinking effort changed from
hightomedium; - thought preservation across turns is now on by default, which improves performance but may increase token usage;
- the model is more expensive than Gemini 3 Flash Preview;
- at launch, Computer Use was not supported on 3.5 Flash; Google recommended staying on Gemini 3 Flash Preview for that.
Related announcements
3.5 Flash also powers Gemini Spark, a personal agent that runs 24/7 to help manage the user's digital life, rolling out first to trusted testers and then to Google AI Ultra subscribers in the US. Google also released Antigravity 2.0, a standalone desktop app for agent-first development.
What this means for businesses
- Cost and speed: "Flash" models with near-"Pro" performance make previously expensive automations viable.
- Test before switching: the new default thinking level and context preservation change cost and behavior. Review prompts and budgets.
- Agents need governance: the more autonomy, the more logs, least-privilege permissions and human review matter.
Sources
- Google, "Gemini 3.5: frontier intelligence with action" (2026-05-19): https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-5/
- TechCrunch: https://techcrunch.com/2026/05/19/with-gemini-3-5-flash-google-bets-its-next-ai-wave-on-agents-not-chatbots/
- Developer guide: https://dev.to/googleai/gemini-35-flash-developer-guide-1i46
- GitHub Changelog: https://github.blog/changelog/2026-05-19-gemini-3-5-flash-is-generally-available-for-github-copilot/
Want to know how to use models like Gemini 3.5 Flash in your business processes? KGA IT offers a free consultation: kga-it.com.
