Google has long positioned Gemini‘s Flash models as the faster, cheaper alternative to its flagship Pro tier. However, that changes with Gemini 3.5 Flash. Announced at I/O 2026, the new model outperforms Gemini 3.1 Pro on coding and agentic benchmarks, with Google claiming that it delivers four times the speed of comparable frontier models, “often at less than half the cost.”
Built for agents
Gemini 3.5 Flash is designed for long-horizon agentic tasks, i.e., workflows that require AI to plan, build, and iterate across multiple steps. Google says it can handle tasks that previously took developers days or auditors weeks, in a fraction of the time.
Specific benchmark scores include 76.2% on Terminal-bench 2.1, 1656 Elo on GDPval-AA, and 83.6% on MCP Atlas. The model also scores 84.2% on CharXiv Reasoning, a multimodal understanding benchmark.
The model works with Google’s Antigravity, an agent-first development platform, to deploy multiple subagents in parallel, enabling it to tackle more demanding workloads. On the consumer side, 3.5 Flash is now the default model powering the Gemini app and AI Mode in Google Search. It also powers Gemini Spark, a new personal AI agent Google announced at I/O that runs around the clock to take action on a user’s behalf. Google is rolling Spark out to trusted testers now, with a broader beta planned for Google AI Ultra subscribers in the US next week.
Rollout and availability
Gemini 3.5 Flash is available now globally in the Gemini app and AI Mode in Search for consumers, and through Google AI Studio, the Gemini API, and Android Studio for developers. Enterprise customers can access it via both the Gemini Enterprise Agent Platform and Gemini Enterprise. Google has also confirmed that Gemini 3.5 Pro is in internal testing and expected to roll out next month.
The release puts agentic capability at the center of Google’s AI roadmap. Rather than competing purely on which model can answer questions more accurately, Google is betting that the next stage of AI is about models that can take actions like booking appointments, writing and running code, managing workflows, and more, all with minimal input from the user. With Gemini 3.5 Pro due next month, it’ll be great to see how far Google plans to push that vision.
