Google ships Gemini 3.5 Flash at I/O; the Pro stays unreleased

At Google I/O on 19 May 2026, Google launched Gemini 3.5 Flash and made it the default model for the consumer Gemini app and AI Mode in Search globally, also shipping it in the Gemini API — all simultaneously. The announcement blog: 'Gemini 3.5: frontier intelligence with action.' In the same announcement, Google also named Gemini 3.5 Pro, its next, more powerful tier — but did not release it. Google said Pro is 'already being used internally' and that it would 'roll it out next month.' No benchmark, context window or price was disclosed for 3.5 Pro.

The capability claim. Google positions Flash as beating Gemini 3.1 Pro — the prior generation's flagship — on coding and agentic benchmarks. The headline figures: 76.2% on Terminal-Bench 2.1, 1656 Elo on GDPval-AA, 83.6% on MCP Atlas, 84.2% on CharXiv Reasoning. Those are Google's own numbers; independent replication from Artificial Analysis and LMSYS is still pending — attribute rather than assert. The price. Official API pricing is $1.50/M input and $9/M output (cached input $0.15/M). Google frames this as 'less than half the cost of other frontier models' — a comparison to flagships, not to the prior Flash. Verify on Google's rate card before budgeting.

What shipped vs what didn't — at a glance:

Verify pricing on Google's current rate card before budgeting at scale.

Google's distribution advantage is the real story here. Making Flash the default in AI Mode in Search — reportedly past a billion monthly users — means the new model is already in front of a vast audience with no opt-in required. That scale of deployment on day one is something only Google can pull off. The capability claims are Google's own; independent evaluation from Artificial Analysis and LMSYS will follow within weeks — check those numbers before treating 76.2% as settled. On Gemini 3.5 Pro: the clock is ticking on the 'next month' promise. Watch for the actual release and the first hard data — Google disclosed no benchmark, context window or price for Pro, so its real capability is unknown until it ships. The gap between the announced model and the shipped model is often where the real story lives.

For developers. Flash's pricing is the most actionable detail in the announcement: $1.50/M input, $9/M output, $0.15/M cached input. Google's 'less than half the cost of other frontier models' is a comparison to flagships, not a low absolute price — and at $9/M output, token-heavy agentic apps will feel it. Check the published rate card against your usage before your next billing cycle. For new projects evaluating which model to build on: Flash is the capable, widely-deployed option; Pro is the one to watch for reasoning-heavy workflows once it actually ships with real numbers.