Google I/O 2026: Gemini 3.5 Flash GA, Antigravity 2.0 + CLI + SDK, and Managed Agents in the Gemini API
At Google I/O 2026 (May 20, 2026), Google made Gemini 3.5 Flash generally available, expanded Antigravity into a desktop app, CLI, and SDK, and launched Managed Agents in the Gemini API for sandboxed agent execution.
Editorial Team
The AI Coding Tools Directory editorial team researches and reviews AI-powered development tools to help developers find the best solutions for their workflows.
Google’s I/O 2026 (May 20, 2026) leaned heavily on developer-facing AI releases. The three changes that matter most for teams building or evaluating AI coding tools: Gemini 3.5 Flash going GA, Antigravity expanding into a desktop app + CLI + SDK, and Managed Agents landing in the Gemini API. (https://blog.google/innovation-and-ai/technology/ai/google-io-2026-all-our-announcements/)
If you’ve been tracking Google Antigravity since its 1.22.2 release or comparing against the Gemini API May 2026 updates, this is the biggest single-day delta of the year.
Agentic development platform powered by Gemini for autonomous coding
Gemini 3.5 Flash is GA
Google released Gemini 3.5 Flash as the first model in its new 3.5 series. It’s available through Google Antigravity, the Gemini API, Google AI Studio, and Android Studio. (https://blog.google/innovation-and-ai/technology/ai/google-io-2026-all-our-announcements/)
On benchmarks, Google says 3.5 Flash outperforms Gemini 3.1 Pro on:
- Terminal-Bench 2.1: 76.2%
- GDPval-AA: 1656 Elo
- MCP Atlas: 83.6%
Google’s positioning: “What used to take a developer days or an auditor weeks, 3.5 Flash can now help complete in a fraction of the time, often at less than half the cost of other frontier models.” (https://blog.google/innovation-and-ai/technology/ai/google-io-2026-all-our-announcements/)
For builders, the practical read is that a Flash-class model is now competitive on agentic coding workloads where Pro-class capacity used to be required.
Antigravity expands into a desktop app, CLI, and SDK
Antigravity evolved from a single product into a platform with three interfaces: (https://blog.google/innovation-and-ai/technology/ai/google-io-2026-all-our-announcements/)
- Antigravity 2.0 — a desktop application for parallel multi-agent orchestration.
- Antigravity CLI — terminal-based agent creation without a graphical interface.
- Antigravity SDK — “programmatic access to the same agent harness powering Google’s products that’s co-optimized for our Gemini models.”
Google also added native voice support for Gemini audio models in Antigravity, plus tighter integrations with Android, Firebase, and Google AI Studio. (https://blog.google/innovation-and-ai/technology/ai/google-io-2026-all-our-announcements/)
For teams already on Antigravity, the CLI and SDK are the more durable announcements — they let you build Antigravity into pipelines, agents, and other products instead of staying inside the IDE.
Managed Agents in the Gemini API
Google introduced Managed Agents in the Gemini API: a single API call provisions a remote environment where agents can: (https://blog.google/innovation-and-ai/technology/ai/google-io-2026-all-our-announcements/)
- Reason, plan, and call tools using the agent harness.
- Execute code and manage files in an isolated sandbox.
- Browse the web to fetch and process live data.
Custom agent extension is supported through markdown files — specifically AGENTS.md and SKILL.md. That’s a notable convergence: those filenames are already familiar from the broader agent-tooling ecosystem, and Google adopting them lowers the friction to port custom agents across platforms.
Pricing and plan updates
Two pricing notes from the same keynote: (https://blog.google/innovation-and-ai/technology/ai/google-io-2026-all-our-announcements/)
- New $100/mo AI Ultra plan with 5× higher usage limits and 20TB of cloud storage.
- Google AI Pro now includes YouTube Premium Lite (an $8.99/mo bundle value).
- AI Studio offers the first two app deployments to Google Cloud at no cost.
Why this matters
Three takeaways for teams building or evaluating AI coding stacks:
- A Flash-class model is now agent-grade. With 3.5 Flash outperforming 3.1 Pro on Terminal-Bench 2.1 at lower cost, the cost ceiling for serious agent workloads drops.
- Antigravity is now a platform. The SDK + CLI make it integratable into existing build/agent pipelines rather than only an IDE.
- Managed Agents shorten the “stand up a sandboxed agent” path from weeks of infra work to a single API call — and align on the same
AGENTS.md/SKILL.mdfiles other agent ecosystems are converging on.
Sources
- Google Blog — “Google I/O 2026: all our announcements”: https://blog.google/innovation-and-ai/technology/ai/google-io-2026-all-our-announcements/
Tools Mentioned in This Article
Free Resource
2026 AI Coding Tools Comparison Chart
Side-by-side comparison of features, pricing, and capabilities for every major AI coding tool.
No spam, unsubscribe anytime.
Frequently Asked Questions
What is Google I/O 2026: Gemini 3.5 Flash GA, Antigravity 2.0 + CLI + SDK, and Managed Agents in the Gemini API?
Related Articles
Cursor 3.9: One customization page for plugins, skills, MCPs, and a marketplace leaderboard
Cursor 3.9 unifies plugins, skills, MCPs, subagents, rules, commands, and hooks on a single customization page, adds custom MCP support, a marketplace leaderboard, plugin canvases, and more marketplace imports.
Read more →Product UpdatesClaude Fable 5 and Mythos 5: Anthropic launches a new frontier model with safeguards and trusted access
Anthropic announced Claude Fable 5 for broad availability and Claude Mythos 5 for a restricted trusted-access program, with safety fallbacks to Opus 4.8 and new rollout/pricing details.
Read more →Product UpdatesCursor 3.8: Automations get a /automate skill, Slack + GitHub triggers, and computer use
Cursor 3.8 expands Automations with a /automate skill for local agents, a Slack emoji trigger, five new GitHub triggers, computer use for cloud agents, and memory file management.
Read more →