Add MobiGyaan as a preferred source on GoogleGoogle has officially introduced the Gemini 3.5 family of AI models, beginning with the launch of Gemini 3.5 Flash, a new model focused on delivering frontier-level intelligence with significantly faster execution for agentic workflows, coding, and long-horizon tasks.
According to Google, Gemini 3.5 Flash is designed to combine high intelligence, low latency, faster response generation, and scalable AI task execution. The company says the model delivers up to 4x faster output token generation compared to other frontier AI models while maintaining strong reasoning and coding capabilities.
Gemini 3.5 Flash Focuses on Agentic AI
Google describes Gemini 3.5 Flash as its strongest agentic and coding-focused model so far. The model is optimized for:
- Multi-step workflows
- Long-duration AI tasks
- Coding assistance
- Enterprise automation
- Intelligent agents
Google says developers no longer need to choose between model quality and response speed as Gemini 3.5 Flash aims to deliver both simultaneously.
Benchmark Performance Improvements
According to Google, Gemini 3.5 Flash outperforms Gemini 3.1 Pro across multiple advanced AI benchmarks including:
- Terminal-Bench 2.1
- GDPval-AA
- MCP Atlas
- CharXiv Reasoning
The company positions the model in the “Top-right quadrant” of the Artificial Analysis index, representing high intelligence and exceptional speed.
Designed for Long-Horizon Workflows
Google says Gemini 3.5 Flash is specifically built for long-horizon agentic tasks where AI systems must plan, execute, iterate, and maintain context across multiple steps. Potential use cases include:
- Application development
- Large-scale codebase maintenance
- Financial document preparation
- Workflow automation
- Research assistance
Antigravity Integration & Collaborative Subagents
The model also works alongside Google’s updated antigravity harness to enable collaborative AI subagents that can handle complex workflows at scale. Google says the system can:
- Reliably execute supervised multi-step tasks
- Coordinate multiple AI agents together
- Maintain high reasoning performance during execution
Example Showcased by Google
Gemini 3.5 Flash was demonstrated:
- Automatically renaming and categorizing unstructured assets
- Using dynamic criteria and multi-step workflows
Improved Multimodal & UI generation Capabilities
Gemini 3.5 Flash also expands on Gemini’s multimodal foundation. The model can generate:
- Richer web interfaces
- Interactive graphics
- Dynamic visual experiences
- Animated research content
Google demonstrated examples, including:
- Interactive visual explanations
- Automated content categorization
- Dynamic generative UI systems
Powering Gemini Spark & Search AI Mode
Google confirmed that Gemini 3.5 Flash is now the default model for the Gemini app globally, and the default model for AI Mode in Google Search. The model is also powering interactive information agents, and dynamic generative UI experiences inside Search.
Gemini Spark Integration
The newly announced Gemini Spark personal AI agent also runs on Gemini 3.5 Flash. Google says Spark can operate continuously, run 24/7 in the background, and take supervised actions on behalf of users. Gemini Spark is rolling out to trusted testers today and coming to Google AI Ultra subscribers in the US next week.
Commenting on the announcement of the Gemini 3.5 family, executives at Google DeepMind said, “Today, we’re introducing Gemini 3.5, our latest family of models combining frontier intelligence with action. This represents a major leap forward in building more capable, intelligent agents. We’re kicking off the series by releasing 3.5 Flash. It delivers frontier performance for agents and coding, excelling at complex long-horizon tasks that deliver real-world utility.”
Availability
Gemini 3.5 Flash is now available across:
- Gemini app
- AI Mode in Search
- Google Antigravity
- Gemini API in Google AI Studio
- Android Studio
- Gemini Enterprise
- Gemini Enterprise Agent Platform
Google also confirmed:
- Gemini 3.5 Pro is already being used internally
- Public rollout begins next month
