Google launches Gemini 3.5 Flash GA and Managed Agents in the Gemini API in public preview on May 19

Google released Gemini 3.5 Flash as a generally available model on May 19, 2026, and simultaneously launched Managed Agents in the Gemini API in public preview — a hosted agent execution service that places Google in direct competition with Anthropic's Managed Agents offering.

What's new

Gemini 3.5 Flash (GA)

• Released as the generally available version of Gemini 3.5 Flash, described as Google's "most intelligent model for sustained frontier performance on agentic and coding tasks"
• Replaces the Gemini 2.0 Flash family, which Google shut down on June 1 and redirected developers to Gemini 3.5 Flash
• Priced at $1.50 per million input tokens and $9.00 per million output tokens on the standard paid tier

Managed Agents (public preview)

• Developers can now build stateful, autonomous agents that run in "secure, isolated Google-hosted Linux sandbox environments"
• Agents are persistent — they maintain state across turns and support long-running autonomous tasks
• Accessible through the standard Gemini API without separate infrastructure provisioning

Context

Both announcements were part of a broader wave of Gemini releases at and around Google I/O in May 2026, which also included Gemini Omni and the Gemini for Science initiative. The Gemini 2.0 Flash family was subsequently deprecated on June 1, with Google directing developers to Gemini 3.5 Flash as the production replacement.

The Managed Agents launch puts Google's hosted agent runtime in direct competition with Anthropic's Managed Agents for the Claude API. Both services offer sandboxed, hosted execution environments for autonomous agents, removing the need for developers to provision and secure their own compute for agentic workloads.

Why it matters

The simultaneous launch of a flagship model and a hosted agent runtime signals how Google views the next generation of developer use cases: not isolated API calls but agents running extended, stateful tasks in cloud environments.

For enterprises evaluating AI infrastructure, Gemini Managed Agents vs. Anthropic Managed Agents vs. OpenAI's sandboxed Codex environments is now a live competitive comparison — three major providers have converged on hosted agent runtimes as a core API-layer feature within roughly the same window.

Gemini 3.5 Flash at $1.50/$9.00 per million tokens positions it as a capable-but-affordable frontier option — cheaper than Gemini 3.1 Pro Preview ($2.00/$12.00) but substantially above Gemini 3.1 Flash-Lite ($0.25/$1.50), which targets high-scale, cost-sensitive workloads.