Google launched Gemini 4 Hydrogen on 12 November 2026 as the fast, everyday model that replaces Gemini 3.8 Flash. It is now in the Gemini app and Google Search for free users. For developers, FourWeekMBA and Unite.AI put its price at twice what 3.8 Flash costs today. Licenseware and Dawloom say 3.8 Flash rises to the same price on 1 January 2027.
If you build on Flash, you have a budgeting decision to make before the year ends.
Where Gemini 4 Hydrogen shows up
Google's launch post calls Hydrogen its "daily driver for agentic workflows." The post lists all tiers of Google AI Studio, the Gemini API, Gemini for business, the Gemini app and Google Search. Paid Google AI Pro and Ultra subscribers also get it in Antigravity.
Unite.AI adds Gemini in Sheets for eligible subscriptions. Digital Trends mentions Android Studio through the Gemini API, and Crypto Briefing lists Vertex AI.
Free users get it, but not on demand
Free Gemini app users don't get a guaranteed Hydrogen option. Unite.AI reports that they reach it through automatic model selection "when Google considers it appropriate for a prompt."
In Antigravity, Hydrogen is limited to paid AI Pro and Ultra plans. The Orchestrator feature rolls out to AI Ultra first. Enterprise customers get both through Gemini for business.
What changes in Google Search
Google promises three changes:
- Better handling of complex, multi-part questions.
- An improved "query fan-out" that links to a wider range of sites.
- More visual answers, including structured responses, comparison tables and custom tools.
Gemini 4 Hydrogen price vs Gemini 3.8 Flash
Google's blog posts don't give prices. The figures below come from third parties that read Google's Gemini API pricing page, as of 12 November 2026. FourWeekMBA and Unite.AI both give Hydrogen at $1.50 per million input tokens and $7.50 per million output tokens. Licenseware, Dawloom and BenchLM give the same 3.8 Flash schedule.
| Model | Input (per 1M tokens) | Output (per 1M tokens) |
|---|---|---|
| Gemini 4 Hydrogen | $1.50 | $7.50 |
| Gemini 3.8 Flash, through 31 Dec 2026 | $0.75 | $3.75 |
| Gemini 3.8 Flash, from 1 Jan 2027 | $1.50 | $7.50 |
Licenseware, Dawloom and BenchLM say the 3.8 Flash cached input, cache storage, batch and priority rates double as well. Licenseware and Dawloom also say Argon's price doubles on 1 December 2026. Pasquale Pillitteri describes the current 3.8 Flash rate as a "promotional price." He also says Hydrogen's price is flat across the full million-token window.
Licenseware says Vertex AI shows the same dates, and that non-global endpoints cost 10% more. FourWeekMBA and CellCog say Hydrogen has a free tier. Neither gives its limits here.
What developers should do now
- Switch the model ID in your API calls to
gemini-4-hydrogen(per FourWeekMBA). - Choose a thinking level for your test runs. Unite.AI lists four: low, medium, high (the default) and a new xhigh. If you set nothing, you get high. The reports linked here don't name the exact request field or AI Studio control for changing it, so look up the thinking-level setting in Google's Gemini API documentation before you script the comparison.
- Run the same set of real prompts once at each level you might use, then compare the billed output-token count for each run. Dawloom's pricing table says output pricing includes thinking tokens, so a higher level that produces more output tokens costs more per request, even at the same list price.
The context window is one shared pool of 1,048,576 tokens for input, thinking and output, per Unite.AI. Google's "2.5x faster" figure is about speed, not cost. With a list price twice 3.8 Flash's current rate, a cheaper cost per task isn't guaranteed.
None of the pricing write-ups linked above gives a shutdown date for 3.8 Flash.
Google's numbers vs independent ones
Google claims Hydrogen beats 3.8 Flash by a wide margin. Unite.AI reports 57% vs 18.2% on Terminal-Bench 4.1, and more than 5x fewer fabricated claims. These are Google's own benchmarks.
The third-party Artificial Analysis Intelligence Index, as reported by Crypto Briefing, scores Hydrogen 51 at default and 54 at its highest setting. Argon, the larger model launched about four weeks earlier, scores 50. The Decoder says benchmarks suggest Hydrogen makes the pricier Argon obsolete. It also cites an independent test that found Hydrogen "rarely hallucinates but lags on terminal work." Check that before you move terminal-heavy agents over.
The date to watch
Argon's price doubles on 1 December 2026. Gemini 3.8 Flash doubles on 1 January 2027, to the same rate as Hydrogen, so staying on 3.8 Flash saves money only until 31 December, and from January Google says you get a better model for the same price. If you run Flash in production, test Hydrogen on your own workload before 31 December, while 3.8 Flash is still the cheaper fallback.
