Gemini 3.5 Flash release notes / what's new

Gemini 3.5 Flash release notes / what's new

Google DeepMind has introduced Gemini 3.5, a new family of models designed to execute complex, agentic workflows. The first release in this series, Gemini 3.5 Flash, is optimized for high-speed, long-horizon tasks in coding and agentic execution, providing frontier-level performance without the latency typically associated with large flagship models.

Gemini 3.5 Flash Performance and Benchmarks

Gemini 3.5 Flash delivers intelligence comparable to large flagship models while maintaining the speed of the Flash series. It is 4 times faster in terms of output tokens per second than other frontier models.

Key performance benchmarks include:

  • Terminal-Bench 2.1: 76.2%
  • GDPval-AA: 1656 Elo
  • MCP Atlas: 83.6%
  • CharXiv Reasoning (Multimodal Understanding): 84.2%

In these areas, 3.5 Flash outperforms Gemini 3.1 Pro, establishing it as Google's strongest agentic and coding model to date.

Agentic Capabilities and Workflow Automation

Gemini 3.5 Flash is designed for long-horizon agentic tasks that require rapid planning, building, and iteration. By reducing the time required for complex tasks—such as maintaining codebases, developing applications, or preparing financial documents—it can complete work in a fraction of the time and at less than half the cost of other frontier models.

When integrated with the updated Antigravity harness, 3.5 Flash can deploy collaborative subagents to execute multi-step workflows and coding tasks under supervision.

Integration and Real-World Applications

Gemini 3.5 Flash is now the default model for the Gemini app and AI Mode in Google Search globally. It also powers several new features:

  • Gemini Spark: A personal AI agent that runs 24/7 to navigate digital life and take action on behalf of the user. It is currently rolling out to trusted testers, with a Beta release for Google AI Ultra subscribers in the US planned for next week.
  • Search Information Agents: New agents that operate 24/7 to provide more intelligent experiences and dynamic generative UI.
  • Enterprise Automation: Banks and fintechs are using the model to automate multi-week workflows, while data science teams use it to extract insights from complex data environments.

Safety and Safeguards

Gemini 3.5 was developed using the Frontier Safety Framework. The model features strengthened cyber and CBRN (Chemical, Biological, Radiological, and Nuclear) safeguards to reduce the generation of harmful content and minimize false refusals of safe queries. These improvements are driven by advanced safety training and interpretability tools that allow developers to analyze the AI's inner reasoning before a response is generated.

Availability

Gemini 3.5 Flash is generally available through the following platforms:

  • For Developers: Google Antigravity, Gemini API in Google AI Studio, and Android Studio.
  • For Enterprises: Gemini Enterprise Agent Platform and Gemini Enterprise.
  • For General Users: The Gemini app and AI Mode in Search.

Google DeepMind also announced that Gemini 3.5 Pro is currently in internal use and is expected to be released next month.

Sources