What happened
Google used its I/O developer conference on May 19 and 20 to ship the Gemini 3.5 model family and a new enterprise agent platform, three weeks after OpenAI and Anthropic each stood up separate joint ventures built to get their models running inside large companies rather than just shipping better benchmarks. Together the three moves point to the same conclusion: the AI labs now see implementation, not raw model quality, as the constraint on enterprise revenue.
Gemini 3.5 lands with agentic and enterprise features
At I/O, Google released Gemini 3.5 Flash, priced at $1.50 per million input tokens and $9 per million output tokens, about 40% below Gemini 3.1 Pro's $2/$12 rate. Google says the model scores 76.2% on the Terminal-Bench 2.1 coding benchmark while generating output roughly four times faster than comparable frontier models, and supports a context window up to 1 million tokens across text, image, video, audio and PDF input.
Gemini 3.5 Pro was announced alongside it but not shipped. CEO Sundar Pichai told the I/O audience to "give us until next month" for general availability, a date Google has not yet confirmed publicly.
Google also introduced Gemini Omni, a model line that combines Gemini's reasoning with generation across image, audio, video and text, and a new $100-per-month AI Ultra tier aimed at developers and technical leads. On the enterprise side, Google Cloud opened the Gemini Enterprise Agent Platform, including a Managed Agents API that lets a company spin up a custom agent inside a Google-hosted, sandboxed environment with no default access to external systems, networks or credentials until explicitly connected. A companion feature, Gemini Spark, is described as a personal AI agent for Workspace and Gemini Enterprise users that requires explicit approval before taking actions such as sending email or editing records, and connects natively to Microsoft SharePoint, OneDrive and ServiceNow rather than requiring an all-Google stack. Gemini Spark is live in Gemini Enterprise now; a Workspace preview is planned for summer 2026.
OpenAI and Anthropic each fund a deployment arm
The I/O announcements followed two enterprise-deployment deals struck days apart in early May. On May 4, OpenAI finalized the OpenAI Deployment Company, a majority-owned subsidiary built to embed OpenAI engineers directly inside client organizations on complex AI projects, a model closer to Palantir's forward-deployed-engineer approach than to shipping software off the shelf. The venture launched with more than $4 billion raised at a $10 billion valuation, backed by 19 firms including TPG, Goldman Sachs, SoftBank, Capgemini and McKinsey & Company. To staff it immediately, OpenAI agreed to acquire Tomoro, a UK-based applied AI consulting firm whose past deployments include Tesco, Virgin Atlantic and Supercell.
The same week, Anthropic finalized a separate $1.5 billion joint venture with Blackstone, Hellman & Friedman and Goldman Sachs, later named Ode, to embed Claude inside companies owned by participating private equity firms. Anthropic, Blackstone and Hellman & Friedman each committed roughly $300 million as anchor partners, Goldman Sachs committed about $150 million, and General Atlantic, Apollo Global Management, Leonard Green, GIC and Sequoia Capital rounded out the capital table. The venture's stated purpose is to handle both the advisory and hands-on implementation work, redesigning workflows and operationalizing the model rather than just licensing it, that PE-owned portfolio companies have struggled to do on their own.
What ties the three together
None of this changes what the underlying models can do. What changes is who does the work of getting a model into production: data integration, workflow redesign, governance and the unglamorous parts of enterprise IT that a chat interface does not solve by itself. OpenAI and Anthropic are answering that by acquiring or renting the consulting layer outright. Google is answering it by building sandboxed agent infrastructure and governance controls directly into its cloud platform, betting that enterprises will deploy agents themselves if the platform handles isolation, approval gating and system integration for them.
For enterprise buyers, the practical effect is that the choice between these vendors increasingly runs through the deployment model on offer, not just a leaderboard score.
What to watch next
Gemini 3.5 Pro's general-availability date is the near-term marker. Pichai told developers to expect it "next month," which would put a release in the June window; Google has not yet confirmed pricing or a firm date.
Sources: Everything Google announced at I/O 2026 (9to5Google) · Google AI updates, May 2026 (Google Blog) · Introducing Gemini Enterprise Agent Platform (Google Cloud Blog) · Gemini 3.5 Flash Released at I/O 2026: pricing (TokenMix) · OpenAI Finalizes $10 Billion Joint Venture With PE Firms to Deploy AI (Bloomberg) · OpenAI launches the OpenAI Deployment Company (OpenAI) · Anthropic teams with Goldman, Blackstone and others on $1.5 billion AI venture (CNBC) · Anthropic and OpenAI are both launching joint ventures for enterprise AI services (TechCrunch)