Google has introduced Gemini 3.5, its latest family of models built to combine frontier intelligence with the ability to take action, kicking off the series with the general release of Gemini 3.5 Flash. Announced at Google I/O 2026, the model is positioned as the company's strongest agentic and coding model yet and is available immediately to billions of users worldwide.
What Google announced
Gemini 3.5 Flash is the first model in the new 3.5 line, with Google describing it as a major step toward more capable, autonomous AI agents. The company said it worked closely with industry partners to identify where complexity and manual effort arise in real workflows, then tuned the model to handle long-horizon tasks that previously took developers days or auditors weeks. Google added that it is also testing Gemini 3.5 Pro internally and plans to roll it out the following month.
Performance and speed
According to Google, 3.5 Flash delivers intelligence rivaling larger flagship models while running at the speeds expected from the Flash line. The company reported that it outperforms Gemini 3.1 Pro on coding and agentic benchmarks including Terminal-Bench 2.1 (76.2%), GDPval-AA (1656 Elo) and MCP Atlas (83.6%), and leads on multimodal understanding with 84.2% on CharXiv Reasoning. Google said the model produces output roughly four times faster than other frontier models, often at less than half the cost. This pace of large language model competition mirrors moves elsewhere in the industry, such as Alibaba's new AI chip built for its Qwen models.
Enterprise and developer adoption
Google highlighted early deployments across banking, fintech and data teams. It said Shopify is running parallel subagents for merchant growth forecasts, Macquarie Bank is piloting the model to reason over 100-plus-page documents during customer onboarding, and Salesforce is integrating 3.5 Flash into its Agentforce platform. Ramp, Xero and Databricks were also named as partners using the model for invoice processing, tax-form workflows and large-scale data diagnostics. The release builds on Google's broader push into agentic AI, an area it has explored in robotics through efforts like its Gemini integration with Boston Dynamics' Spot robot.
Availability and safeguards
Gemini 3.5 Flash is now the default model in the Gemini app and AI Mode in Google Search, and is available to developers through Google Antigravity and the Gemini API in Google AI Studio and Android Studio, as well as to enterprises via the Gemini Enterprise Agent Platform. Google said the model was developed under its Frontier Safety Framework, with strengthened cyber and CBRN safeguards and interpretability tools designed to examine the model's reasoning before it responds. The rapid expansion of capable AI systems has drawn growing policy attention, including the European Parliament's move to create a permanent AI and robotics committee.
Reporting based on coverage from Google (The Keyword blog).