Google Is Testing ‘Hydrogen,’ a Fast New Gemini 4 Model

Google is internally testing a new Gemini 4 model, one that appears designed to bring the company's latest generation of AI to a faster, cheaper tier.
Google Gemini
Google Is Testing ‘Hydrogen,’ a Fast New Gemini 4 Model

Google is internally testing a new Gemini 4 model, one that appears designed to bring the company’s latest generation of AI to a faster, cheaper tier.

The model is known internally as Hydrogen, according to Business Insider, which viewed an internal message about it. Hugh Langley, a senior correspondent for the outlet, reports that a Google engineer described the model to staff this week as “very capable, fast and efficient.”

Google employees can access Hydrogen through Jetski, the company’s internal development platform, where new models often appear ahead of a public release. One employee who has used the model told Business Insider it is similar to Google’s previous Flash models, the workhorse line the company positions as faster and more cost-effective than its top-tier offerings. Google declined to comment on the record.

Filling Out the Gemini 4 Lineup

If Hydrogen is indeed a Flash-class model, it would fill an obvious gap in Google’s Gemini 4 family.

Google announced Gemini 4 Argon on Sept. 30, initially limiting access to cybersecurity defenders before opening it to developers and Google AI Pro and Ultra subscribers on Oct. 13. The company followed with Gemini 4 Carbon on Oct. 28, describing it in a blog post as:

our strongest model with state-of-the-art cybersecurity and long horizon capabilities.

Carbon followed the same staged path as Argon, starting with select cyber defenders before Google opened it to paying developers and Google AI Ultra subscribers on Tuesday.

The common thread is that both models are available only to paying customers. In the Gemini API, Argon and Carbon cost $2 per million input tokens and $10 per million output tokens under introductory pricing that runs through Nov. 30. Starting Dec. 1, those rates double to $4 and $20, respectively.

In contrast, Google’s current Flash model, Gemini 3.8 Flash, launched in early September at $0.75 per million input tokens and $3.75 per million output tokens. What’s more, Google offers Flash models free of charge in the Gemini app. According to Business Insider, it’s unclear whether Hydrogen will follow suit or be reserved for paying users.

Why Flash Matters

While flagship models like Carbon grab the headlines and top the benchmark charts, Flash models arguably matter more to the average user. They power the free tier of the Gemini app, handle the bulk of everyday queries and give developers an affordable option for high-volume workloads where speed matters more than raw reasoning power.

In short, a Gemini 4-class Flash model would put much of the new generation’s improvements in the hands of the very people least likely to pay for Argon or Carbon.

The name certainly fits the role. As Business Insider points out, hydrogen is the lightest element on the periodic table, making it an apt codename for a lightweight model in a family that already includes Argon and Carbon.

Leakers Got There First

Business Insider’s report lends credibility to chatter that has been circulating among AI model watchers for the past couple of weeks. As OfficeChai reported, several accounts on X that track unreleased models began posting about a Flash-tier Gemini 4 model codenamed Hydrogen in late October. Some of those posts claimed Google was testing five different Hydrogen checkpoints, with output speeds ranging from roughly 300 to 500 tokens per second.

It’s important to note, however, that Google has not confirmed any of those details, and the claims have not been independently verified. Nor has a Hydrogen or Gemini 4 Flash model appeared in Google’s public developer documentation.

Google’s Comeback Continues

The report comes at an interesting time for Google. Earlier this year, the company delayed and ultimately canceled Gemini 3.5 Pro, leading many to conclude it had fallen behind in the race to build the most advanced AI models. The rapid-fire releases of Argon and Carbon have gone a long way toward changing that narrative.

The competition is not standing still, either. Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on Monday, touting improvements in coding, reasoning and vision.

Google has not announced when, or even if, Hydrogen will be released to the public. Nonetheless, if the company follows the pattern it established with Argon and Carbon, Hydrogen may not stay inside Google’s walls for long. When it does arrive, it could prove to be the most widely used Gemini 4 model of them all.

Subscribe for Updates

GenAIPro Newsletter

News, updates and trends in generative AI for the Tech and AI leaders and architects.

By signing up for our newsletter you agree to receive content related to ientry.com / webpronews.com and our affiliate partners. For additional information refer to our terms of service.

Notice an error?

Help us improve our content by reporting any issues you find.

Get the WebProNews newsletter delivered to your inbox

Get the free daily newsletter read by decision makers

Subscribe
Advertise with Us

Ready to get started?

Get our media kit

Advertise with Us