Google’s reported “Hydrogen” Gemini 4 model targets speed and cost
Business Insider says Google is testing a Gemini 4 model called Hydrogen through Jetski. Its price, benchmarks and release date remain unknown to the public.

Why it matters
- Business Insider reports Google is testing a Gemini 4 model codenamed Hydrogen through its Jetski platform.
- Google announced Gemini 4 Argon on Sept. 30 and expanded access Oct. 13; Gemini 4 Carbon followed Oct. 28.
- Argon and Carbon introductory API pricing is $2 per million input tokens and $10 per million output tokens through Nov. 30, then $4 and $20.
- Gemini 3.8 Flash launched in early September at $0.75 per million input tokens and $3.75 per million output tokens.
- Unverified posts claimed five Hydrogen checkpoints and output speeds of roughly 300 to 500 tokens per second.
By James Calloway
Google is internally testing a Gemini 4 model codenamed Hydrogen through Jetski, its internal development platform, according to Business Insider. Business Insider senior correspondent Hugh Langley said the outlet viewed an internal message about the model.
He reported that a Google engineer described it to staff this week as “very capable, fast and efficient.” The report did not name the engineer or attach supporting benchmarks.
Google employees can access Hydrogen through Jetski, where new models often appear before a public release, Business Insider reported. One employee who used the model said it resembles Google’s earlier Flash models, the workhorse line Google positions as faster and more cost-effective than its top-tier offerings. Google declined to comment on the record.
WebProNews frames the stakes as distribution and cost, not merely benchmark leadership. Google lists introductory API (application programming interface) prices of $2 per million input tokens (units of text processed by a model) and $10 per million output tokens for Argon and Carbon. Gemini 3.8 Flash costs $0.75 and $3.75 on the same basis.
What does the internal report establish?
On the public record, Google has said nothing beyond declining comment. Business Insider documented Hydrogen’s existence through an internal message, while employees reported using it on Jetski. That supports the claim that Google has an experimental model in circulation.
The report does not publish benchmarks, model size, context limits, training details, pricing, a release date or technical documentation. It also does not say whether Hydrogen is a final Flash variant, a test build or a broader successor. Nor does it establish whether Hydrogen will remain the public name.
WebProNews says a Flash-class Hydrogen would fill an obvious gap in the Gemini 4 family. That is a lineup judgment, not a confirmation of specifications.
Business Insider’s internal message therefore establishes more than anonymous chatter, but less than a launch commitment. The engineer’s quoted assessment remains a vendor claim without independent testing.
Where does Hydrogen fit in the Gemini 4 lineup?
Google announced Gemini 4 Argon on Sept. 30, initially limiting access to cybersecurity defenders. Google then opened it to developers and Google AI Pro and Ultra subscribers on Oct. 13.
The company followed with Gemini 4 Carbon on Oct. 28. Google wrote in a blog post: “our strongest model with state-of-the-art cybersecurity and long horizon capabilities.” Long horizon means extended, multi-step tasks. Carbon repeated Argon’s staged path, starting with select cyber defenders before reaching paying developers and Google AI Ultra subscribers on Tuesday.
Both models remain restricted to paying customers. Their $2 per-million-input-token and $10 per-million-output-token rates count as introductory pricing through Nov. 30. Starting Dec. 1, Google will double them to $4 and $20.
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. Google also offers Flash models free in the Gemini app. Business Insider could not determine whether Hydrogen will follow that free model or remain a paid product.
Why could Flash matter more than the flagships?
WebProNews argues that Flash models may matter more to ordinary users than headline flagships, even when Carbon tops benchmark charts. Flash powers the Gemini app’s free tier and handles most everyday queries.
That position gives Google a cheap route to volume. Developers can use Flash for high-volume workloads (large numbers of automated requests) where response speed matters more than maximum reasoning power, according to the report.
WebProNews says a Gemini 4-class Flash model would put new-generation improvements in front of people least likely to buy Argon or Carbon. The outlet also noted that hydrogen is the lightest element on the periodic table, making the codename fit a lightweight model in a family that already includes Argon and Carbon.
The naming rationale is suggestive, but it supplies no performance evidence. The source provides no confirmed speed, quality or cost figure for Hydrogen, so the codename cannot substitute for independent benchmarks.
What do the leaked specifications prove?
Business Insider’s report lends some credibility to chatter that had circulated among AI model watchers for the previous couple of weeks. OfficeChai reported that several accounts on X tracking unreleased models began posting about a Flash-tier Gemini 4 model called Hydrogen in late October.
OfficeChai relayed posts claiming Google was testing five Hydrogen checkpoints (saved model versions). The posts also claimed output speeds of roughly 300 to 500 tokens per second, rather than citing Google documentation or a reproducible test.
Google has not confirmed those details, and no one has independently verified them. Neither Hydrogen nor a Gemini 4 Flash model appears in Google’s public developer documentation, according to Business Insider.
Business Insider’s internal message strengthens the case that Hydrogen exists, but it does not corroborate the checkpoint count or speed range. A credible leak and a verified specification are different things.
How does Hydrogen fit Google’s model strategy?
WebProNews noted that Google delayed and ultimately canceled Gemini 3.5 Pro earlier this year. The report said many observers concluded that the cancellation left Google behind in the race to build the most advanced AI models.
The launches of Argon and Carbon in quick succession have partly changed that narrative, WebProNews said. Competition has not paused: Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on Monday, citing improvements in coding, reasoning and vision.
WebProNews says Hydrogen may not remain internal for long if Google repeats the staged rollout used for Argon and Carbon. The outlet also argues that it could become the most widely used Gemini 4 model.
That expectation remains a forecast, not a launch signal. Business Insider reports no public developer entry and no release timetable for Hydrogen.
What would make Hydrogen commercially important?
Commercial terms will determine whether Hydrogen changes the Gemini 4 market or simply extends its premium tier. If Google gives it the free-app access used by Flash, it could carry Gemini 4 improvements to users who will not buy Argon or Carbon.
If Google reserves it for paying subscribers, Hydrogen would sit closer to Argon and Carbon on cost, with less chance of becoming the family’s default model. Business Insider said pricing, free availability and a public release date all remain unresolved.
The next useful disclosures are straightforward: launch timing, API prices, free-app access, model documentation and independently repeatable benchmarks. Until those appear, Hydrogen is a credible internal project, not a verified commercial product.
If Google gives Hydrogen the free-app treatment used by Flash, Gemini 4’s improvements could reach users without a paid subscription; if it keeps the paid pattern of Argon and Carbon, the model would arrive as another premium tier.
Source: webpronews.com
More from James Calloway
Show full bio
News editor covering people, launches and day-to-day practice in artificial intelligence and machine learning at AI In Context.
538 articles
Related articles
- Google Cuts Gemini Flash Pricing in Half With 3.7 Release
- Google Launches Gemini 4 Hydrogen at $1.50 Per Million Input Tokens
- Google Launches Gemini 3.1 Flash-Lite Starting at $0.25 Per Million Tokens
- Google Launches Gemini 3 Pro at $2/Million Input Tokens
- Google Ships Gemini 3.8 Flash and a Cybersecurity-Only Variant