Gemini 4 Hydrogen is live globally — and it’s already reshaping who gets what, for how much, and when. Google rolled it out on November 12 across the Gemini app, Google Search, and the Gemini API. It replaces Gemini 3.8 Flash as the default model for everyday use, the first to drop “Flash” from its name.
Free users don’t get a toggle. They get Hydrogen only when Google’s auto model selection decides it’s needed, usually for prompts demanding more reasoning. Most still go to Flash-Lite. Turn off auto selection?
Every reply defaults to Flash-Lite, no exceptions. That’s been true since October, and it’s unchanged. You can verify which model generated a response: click the menu under any reply in gemini.google.com → Show thinking steps → scroll to the bottom. The model name appears there.
Settings > Usage limits holds the auto-selection switch. Free tier users see no model picker at all. Paying users do, but the pricing tiers just shifted hard. Google AI Plus ($4.99/month) now headlines Gemini 4 Hydrogen as its flagship model, with double the usage limits of the free tier.
Gemini 3.1 Pro has vanished from the Plus column. It’s gone from the plans page entirely. Instead, Google AI Pro starts at $19.99/month and bundles both Hydrogen and the larger Gemini 4 Argon. An email to Plus subscribers confirmed Pro models would be cut off on November 13.
But Google’s help page still lists Pro under AI Plus, a clear mismatch between marketing copy and backend reality. Developers pay $1.50 per million input tokens and $7.50 per million output tokens. Context window: 1 million tokens. Four thinking levels are available, including a new xhigh setting for deep-dive answers.
In benchmarks, Hydrogen finishes tasks 2.5× faster than 3.8 Flash and hallucinates more than five times less often. On the Artificial Analysis Intelligence Index, it scores 51, one point ahead of Argon. Google says Hydrogen powers multi-part questions and typo-tolerant search for all users worldwide. No regional restrictions.
No waitlist. No country-specific rollout delay. It’s here, and the pricing and access rules are already in effect. Next: Google will update its help documentation to match the live plans page.
Until then, the discrepancy stands.
ADFiled by The AI Desk
Models, assistants and the companies and chips behind them, reported from what was released and what was claimed, with the difference kept clear.
More from this desk →
Be the first to comment
Join the argument. No password, just your email or a passkey.