Google used its I/O 2026 developer conference to ship a full-stack agentic AI platform. The company launched Gemini 3.5 Flash, a new frontier model optimized for agents and coding. It debuted Antigravity 2.0, a desktop application for orchestrating parallel agent workflows. And it introduced Managed Agents in the Gemini API, letting developers spin up persistent agent environments with a single API call.

CEO Sundar Pichai framed the announcements as a structural pivot. “We’re firmly in our agentic Gemini era,” he said, citing 3.2 quadrillion tokens processed monthly across Google products, up from 480 trillion a year ago.

Gemini 3.5 Flash: Faster, Cheaper, Agent-Optimized

Gemini 3.5 Flash is the first model in what Google calls its “frontier intelligence with action” family. According to MarkTechPost, the model scores 76.2% on Terminal-Bench 2.1 for coding, 1656 Elo on GDPval-AA for agentic task performance, and 83.6% on MCP Atlas for tool-use reliability. It outperforms Gemini 3.1 Pro across nearly all benchmarks while running four times faster.

Pricing targets high-volume agent workloads: $1.50 per million input tokens, $9.00 per million output tokens, and $0.15 per million cached input tokens. The context window supports 1,048,576 input tokens with dynamic thinking enabled by default, meaning the model auto-allocates more compute for harder problems.

Google says enterprise partners are already running production workloads on 3.5 Flash. Shopify uses it for parallel data analysis, Macquarie Bank is piloting it for customer onboarding over 100+ page documents, and Salesforce is integrating it into Agentforce for multi-turn tool calling with retained context.

Antigravity 2.0: Desktop App, CLI, SDK

The Antigravity 2.0 desktop application acts as a central hub for orchestrating multiple agents running in parallel. It supports dynamic subagents for parallelized workflows, scheduled tasks for background automation, and integrations with Google AI Studio, Android, and Firebase.

Alongside the desktop app, Google released Antigravity CLI for terminal-based development, an SDK for programmatic agent harness access, and enterprise integration through Google Cloud’s Agent Platform. The company is encouraging existing Gemini CLI users to migrate to Antigravity CLI.

A new $100/month AI Ultra subscription plan provides 5x higher usage limits in Antigravity compared to the AI Pro plan. Google offered new subscribers $100 in bonus Antigravity credits through May 25.

Managed Agents in the Gemini API

The most technically significant announcement for developers: Managed Agents in the Gemini API. A single API call now spins up an agent that reasons, uses tools, and executes code inside an isolated Linux container. Files and state persist across follow-up calls, enabling multi-turn sessions without manual environment management.

The managed agents run on Google’s Antigravity agent harness, co-optimized with Gemini 3.5 Flash. Developers can define custom agent behaviors using markdown files and access templates through Google AI Studio.

Gemini Spark: Google’s OpenClaw Competitor

Google also unveiled Gemini Spark, a cloud-based personal AI agent that runs 24/7 in the background. Unlike OpenClaw’s local-hardware approach, Spark operates entirely on cloud virtual machines, continuing to work when a user’s laptop is closed. According to Mashable, Spark has native access to Gmail, Google Docs, and Google Drive for users already in Google’s ecosystem.

Pichai said a beta would be available to AI Ultra subscribers first. Google also introduced Agent Payments Protocol (AP2), which lets users set strict spending limits on what agents can purchase and from which merchants.

The Competitive Timing

The I/O announcements landed as OpenClaw crossed 300,000 GitHub stars. Google’s response covers every layer of the agent stack: a purpose-built model (3.5 Flash), a development platform (Antigravity 2.0), managed infrastructure (API agents), a consumer product (Spark), and a payment guardrail (AP2). The question is whether Google’s cloud-native, subscription-gated approach can match the velocity of an open-source ecosystem that already has 8.5 million developers building on Gemini, according to Pichai’s own figures, plus a growing OpenClaw community outside Google’s walls.