Google Unveils Gemini Spark and Antigravity Tools

Google used its I/O 2026 keynote and product blogs to announce multiple agentic and model updates. Per Sundar Pichai's I/O remarks published on the Google blog, monthly token consumption across Google surfaces rose to over 3.2 quadrillion tokens per month, up from 480 trillion last year. The Google Cloud blog lists a set of offerings for enterprises and developers, naming `Gemini 3.5` (launched with Gemini 3.5 Flash), `Gemini Omni`, Google Antigravity, and enterprise features including Gemini Spark, Managed Agents API on Agent Platform, and an AI security agent called CodeMender.

Per the Google Cloud blog, Gemini 3.5 Flash targets agentic and coding workloads and is reported to outperform prior Gemini releases on benchmarks the company cited (Terminal-Bench 2.1, GDPval-AA, MCP Atlas). Reporting from TechCrunch and The Verge describes `Gemini Spark` as a cloud-hosted, always-on personal agent built from Gemini base models that runs on dedicated Google Cloud VMs and integrates with Gmail, Docs, Sheets, Slides and third-party services via the MCP standard. The Verge and Mashable report that Google said Spark will be able to access local files on macOS through the Gemini app and that users will be able to text and email Spark directly.

Google highlighted Google Antigravity (agentic coding IDE) updates and a desktop app and CLI in reporting by TechCrunch and Google Antigravity's site. The Managed Agents API on Agent Platform was announced for building and running custom agents in Google-hosted environments, per the Google Cloud blog. CodeMender was presented as an AI security agent available through Agent Platform.

Industry-pattern observations: persistent, cloud-hosted personal agents like Gemini Spark follow a recent wave of agentic products from major labs and emphasize provider-side compute. For practitioners, that shifts implementation trade-offs: less local hardware burden but greater reliance on provider-managed VMs, OAuth-style connectors, and hosted data access patterns.

Editorial analysis: Google combining frontier model updates (Gemini 3.5) with agent runtimes (Agent Platform, Managed Agents API) and developer tooling (Antigravity, CodeMender) continues the industry trend toward integrated stacks where models, runtimes, and connectors ship together. Observers quoted in TechCrunch and The Verge highlight the competitive angle against other agent products and note Google's advantage in Workspace and personal data signals.

For practitioners: monitor the Managed Agents API and Antigravity toolchain for integration and security primitives, evaluate how MCP connectors manage least-privilege access to third-party services, and watch availability windows, TechCrunch reports Spark is in internal testing and expected to become available to Google AI Ultra subscribers next week. Also track the security posture of CodeMender and hosted-agent defaults, since cloud-hosted persistent agents change attack surfaces and data governance requirements.

TechCrunch attributes a direct remark from CEO Sundar Pichai: "It's your personal AI agent that helps you navigate your digital life, taking action on your behalf and under your direction." The Verge cites Josh Woodward, VP of Google Labs, saying, "When you use it, it almost feels like you're tossing things over your shoulder, Spark's catching them, and gets the job done."