
What happened
Google launched Gemini 4 Hydrogen with immediate access and no paid-tier restriction, replacing Gemini 3.8 Flash across its products; Google says it runs 2.5 times faster than Flash 3.8 even in high thinking mode and cuts fabricated claims by more than 5x.
Why it matters
The relaxed rollout contrasts with the staged, cybersecurity-first releases of Argon and Carbon, so the Flash-level users most people in Google's ecosystem rely on get the upgrade right away, with code-generation gains that more than triple 3.8 Flash's Terminal-Bench score.
WHO IT HITSDevelopers and everyday users of Google's cheaper Flash-tier models get the upgrade immediately rather than waiting behind paid-tier or security-review gates, and teams relying on code generation may see materially different benchmark results.
Summaries like this, every morning.
Google's elemental naming scheme has moved away from the descriptive Pro/Flash labels, but Hydrogen's role is unchanged: it is the faster, more economical model that most people in Google's ecosystem actually use. What is different this time is the release process. The recent Argon and Carbon models were staged, starting with cybersecurity researchers and, even at general availability, limited to users paying for Google's most expensive AI subscription.
Hydrogen skips those gates. Access began immediately, and the model has already replaced Gemini 3.8 Flash across the company's products. Google's stated gains are speed and reliability: it runs in one of four thinking levels (low, medium, high, and the new xhigh), and even in high mode it is 2.5 times faster than Flash 3.8. Google's analysis also puts the reduction in fabricated claims at more than 5x.
On code generation, Google says Hydrogen more than triples 3.8 Flash's score on Terminal-Bench, and in some cases it trades blows with the higher-tier Argon — scoring 54 on the Artificial Analysis Intelligence Index in xhigh mode versus 50 for Argon. That positioning suggests the cheaper tier is narrowing the gap with Google's premium model, though the comparison rests on Google's own analysis and on benchmark scores rather than independent testing.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Omaha companies, including Union Pacific with its 'Rita' tool and others using Copilot, shared how they are ad…

OpenAI agreed to acquire Odessia, an AI trip-planning engine that sits between travelers and booking platforms…

ServiceNow launched its Autonomous CRM Engine, which uses AI to analyze existing CRM setups and automate workf…

Google and Samsung's new smart glasses, built on Android XR with Gemini, redact faces and ambient audio on-dev…

Okinawa Toyota Motor ran a demonstration of its "XR Ride Bus × AI Guide" in Miyakojima City on the 10th, with…

Pietro Miotti said 48 AI agents combing NASA lidar data flagged a 'major village' in the Maya region; Mexico's…
