Google rolled out Gemini 3.5 Flash on Tuesday at its annual I/O developer conference, positioning the model as the company’s most powerful tool for coding and autonomous AI agents. The announcement underscored a strategic pivot: moving away from chat‑driven interfaces toward AI that can act, plan, and build with little human prompting.
DeepMind’s chief technologist, Koray Kavukcuoglu, told reporters the new model delivers "an incredible combination of quality and low latency." In internal benchmarks, Gemini 3.5 Flash outperformed the earlier 3.1 Pro across coding, multimodal reasoning, and agentic tasks. Kavukcuoglu noted a four‑fold speed edge over other frontier models, while an optimized flash variant runs twelve times faster without sacrificing accuracy. The speed boost is designed for agentic workloads where several AI agents operate concurrently on long‑running projects.
During the live demo, Google engineer Varun Mohan showed agents spawning off to tackle separate components of a full operating system inside Antigravity, the company’s agent‑first development platform. The agents collaborated, merged their outputs, and produced a working OS—all without direct human intervention. Antigravity 2.0, released alongside the model, is a stand‑alone desktop application built around this agentic workflow.
Google says the model is already delivering tangible results for partners. Financial institutions and fintech firms have used Gemini 3.5 Flash to automate multi‑week workflows, while data‑science teams report faster insight extraction from complex datasets. The system can run autonomously for several hours, pausing only when it encounters a decision point that requires human judgment, according to Tulsee Doshi, Google’s senior director of product.
Looking ahead, Google plans to pair the flash model with a forthcoming 3.5 Pro variant. Doshi described 3.5 Pro as the "orchestrator" that will plan and allocate tasks, delegating execution to the flash sub‑agents. Together, the two models aim to balance deep reasoning with brute‑force computational power.
Gemini 3.5 Flash is now the default model in the Gemini app and in AI Mode on Google Search worldwide. The company also announced new agentic capabilities for Search, allowing users to create, customize, and manage AI agents directly on the platform. The same model powers Gemini Spark, a personal AI assistant designed to run 24/7 and help consumers manage their digital lives.
The broader rollout raises safety concerns. Google is currently defending a lawsuit filed by the family of a man who died by suicide after weeks of interacting with Gemini. In response, the company says Gemini 3.5 Flash includes strengthened cyber and CBRN safeguards and is better calibrated to engage with sensitive queries rather than refusing them outright.
Gemini 3.5 Flash is available today through Antigravity, the Gemini API, Gemini Enterprise, the Gemini app, and AI Mode in Search.
Cet article a été rédigé avec l'assistance de l'IA.
News Factory SEO vous aide à automatiser le contenu d'actualités pour votre site.