Google unveiled Gemini 3.5 Flash on Tuesday at its annual I/O developer conference, calling it the company’s strongest model yet for coding and autonomous agents. According to TechCrunch AI, the model can independently execute coding pipelines, manage research projects, and in internal tests built an entire operating system from scratch. This is Google planting a flag: the next wave isn’t about chattier chatbots, it’s about AI that plans, builds, and ships work with minimal human babysitting.
What stands out here is the speed claim. Koray Kavukcuoglu, DeepMind’s chief technologist, told reporters that 3.5 Flash “outperforms our latest frontier model, 3.1 Pro, on nearly all the benchmarks,” including coding, agentic tasks, and multimodal reasoning. He pegged it at 4x faster than other frontier models, with an optimized variant clocking in 12x faster at the same quality. Speed matters because agents fan out, work in parallel, and run for hours. Latency is the tax on every step they take.
What’s new
- Agent-native design. Flash 3.5 was co-developed with Antigravity, Google’s agentic IDE, so agents have a “native environment where they can live, work, and execute,” Kavukcuoglu said. Antigravity 2.0, a standalone desktop app, also launched at I/O.
- Long-running autonomy. The model can run for multiple hours on its own. Tulsee Doshi, Google’s senior director and head of product, said it pauses for human input when it hits a decision point or permission gate.
- Pro plus Flash, working together. When the forthcoming 3.5 Pro lands, Doshi told TechCrunch AI, “3.5 Pro becomes your orchestrator, your planner, and then it actually can leverage Flash to be the various sub-agents.” Pro handles reasoning. Flash handles brute-force tool use.
- Default everywhere. 3.5 Flash is now the default model in the Gemini app and in AI Mode in Search globally. It also powers Gemini Spark, Google’s new always-on personal agent.
Why it matters
For practitioners, this is the clearest signal yet that the major labs are racing to make agentic workflows production-grade, not just demo-grade. Onstage at I/O, Google engineer Varun Mohan showed agents spawning off to build separate components of an operating system inside Antigravity, then merging the work. Google says partners are already using 3.5 Flash to automate multi-week workflows at banks and fintechs, and to surface insights in messy data environments.
That’s a different pitch from the conversational AI era. The competitive frame has shifted from “who has the smartest chatbot” to “whose agents can actually finish a job.” Anthropic, OpenAI, and Google are all chasing the same horizon. Google’s bet is that pairing a fast workhorse model with a planner-orchestrator is the architecture that wins.
The safety question
More capability means more exposure. Google is facing a lawsuit after a man nearly committed a mass casualty event and died by suicide following weeks of chatting with Gemini last year. Handing autonomous agents to a mass consumer audience multiplies that risk. Google says 3.5 Flash has strengthened cyber and CBRN (chemical, biological, radiological, and nuclear) safeguards and is calibrated to engage with sensitive questions rather than refuse them outright. Whether that calibration holds at scale is the open question.
How to get it
Gemini 3.5 Flash is generally available today through Antigravity, the Gemini API, Gemini Enterprise, the Gemini app, and AI Mode in Search. Developers building agentic workflows should test it head-to-head against their current stack, especially on long-running, tool-heavy tasks where the latency advantage compounds.
Full details and demos are available at the original TechCrunch AI report.