Google released Gemini 4 Hydrogen on November 12 as its new fast, lower-cost model for everyday coding and agent work. It replaces the “Flash” name and costs $1.50 per million input tokens and $7.50 per million output tokens. Google says Gemini 4 Hydrogen finishes tasks 2.5 times faster than Gemini 3.8 Flash.
Below: what Gemini 4 Hydrogen costs, who can use it, how Google’s claims compare with independent tests, and what to check. The checklist is The Jolt Journal’s own. The Jolt Journal did not run the model.
Gemini 4 Hydrogen at a glance
| Item | What Google says |
|---|---|
| Release | November 12, 2026 |
| Role | “Daily driver for agentic workflows,” per Google’s announcement |
| Context window | 1 million tokens of input, up to 1 million of output |
| Thinking levels | Low, medium, high (the default) and a new xhigh |
| Inputs | Text, images, audio and video, per VentureBeat |
- Google positions it beside Gemini 4 Carbon and Argon, its larger models for long, deep reasoning.
- The announcement was updated on November 13.
What Gemini 4 Hydrogen costs
Google lists flat pricing across the full context window. VentureBeat added the cached and discount tiers from Google’s pricing.
| Tier | Price per million tokens |
|---|---|
| Input | $1.50 |
| Cached input | $0.075 |
| Output, including reasoning tokens | $7.50 |
| Batch, Flex and Contributor tiers | Half the standard rate |
VentureBeat said Batch targets a 24-hour turnaround, Flex targets one to 15 minutes with limited availability, and Contributor is invite-only and requires letting Google use request data.
Gemini 4 Hydrogen versus other Gemini models on price
VentureBeat’s headline called Gemini 4 Hydrogen “pricier” than its predecessor. The table shows why.
| Model | Input / output per million tokens | Note |
|---|---|---|
| Gemini 3.8 Flash | $0.75 / $3.75 | Introductory rate through December 31 |
| Gemini 3.8 Flash | $1.50 / $7.50 | Scheduled from January 1, 2027 |
| Gemini 4 Hydrogen | $1.50 / $7.50 | Standard rate |
| Gemini 4 Argon and Carbon | $2 / $10 | Introductory; $4 / $20 afterward |
By The Jolt Journal’s arithmetic, a job with 1 million input tokens and 200,000 output tokens costs $3.00 on Gemini 4 Hydrogen at the standard rate. The same job costs $1.50 at the Batch rate, $1.50 on Flash at its introductory price, and $4.00 on Argon or Carbon at theirs.
How fast and smart is Gemini 4 Hydrogen?
The numbers come from two places: Google’s own tests and independent ones. They should be read separately.
| Measure | Result | Source |
|---|---|---|
| Speed versus Gemini 3.8 Flash | 2.5 times faster at the high setting | |
| Terminal-Bench 4.1 | 57%, up from 18.2% for Flash | |
| Intelligence Index, high | 51 (Argon scores 50) | Artificial Analysis, as cited by Google |
| Intelligence Index, xhigh | 54 | Artificial Analysis |
| Output speed, xhigh | 185 tokens per second | Artificial Analysis |
| Time to first answer token, xhigh | About 43 seconds | Artificial Analysis |
The wait to a first answer is the catch. Artificial Analysis put the median for comparable models at about 13.6 seconds, so the xhigh setting is slow to start. It also listed a cost of $3.16 per index task at xhigh.
What partners reported about Gemini 4 Hydrogen
- Google’s partners. Early testers saw runs 2x to 2.8x faster and 12% to 23% lower token use than Flash.
- Box. VentureBeat said Box CEO Aaron Levie posted that accuracy on its complex work test rose from 74% to 79%.
- Sheets. Google said outputs in the Sheets canvas arrive more than 30% faster than before.
These are company claims. They were not tested by The Jolt Journal.
Who gets Gemini 4 Hydrogen
| Audience | Access, per Google |
|---|---|
| Free Gemini app users | Through auto model selection, when Google decides it fits a prompt |
| Gemini app picker | Labeled “Everyday help,” per 9to5Google |
| Google Search | All users globally, as the model behind AI Mode |
| Developers | Gemini API in AI Studio and Android Studio |
| Google Antigravity | Paid Google AI Pro and Ultra subscribers |
| Businesses | Gemini for business; Gemini in Sheets for eligible plans |
The site’s report on Google AI Plus losing Gemini Pro selection covers another change to Google’s plan lineup this week.
How Gemini 4 Hydrogen works with Carbon in Antigravity
Google added an Orchestrator to Antigravity, its coding tool. Per 9to5Google, Carbon can hand routine steps to Gemini 4 Hydrogen and switch models within one session. Google says the pairing gets near-Carbon task success while using 24% less quota and finishing 19% faster, on Terminal-Bench 4.0.
- Antigravity also gets an “Extra” thinking tier beside low, medium and high.
- Access needs a paid Google AI Pro or Ultra plan.
- Google’s figures are for its own setup, not a general result.
In Search, Gemini 4 Hydrogen is the default model for AI Mode, and Google says it handles the hardest questions in AI Overviews. Google said the upgrade understands multi-part and typo-filled queries better.
Safety claims for Gemini 4 Hydrogen
Google listed several behavior results against Flash. They are Google’s own evaluations.
- Overreach in agent tasks down by up to 81%.
- Under-service down by up to 79%.
- Misaligned actions and shortcutting down 12x.
- Up to 2x less likely to bypass permission limits on blocked tasks.
- More than 2.3x more resilient to indirect prompt injection on one benchmark.
Google’s blog gave a 94% score on its internal honesty test. 9to5Google reported 96.7%. The Jolt Journal could not reconcile the two.
A checklist before switching to Gemini 4 Hydrogen
These are The Jolt Journal’s suggestions, not Google’s guidance.
- Test your own task. Benchmarks do not match your data. Run a sample first.
- Watch the default. VentureBeat said the default thinking level is high, where Flash defaulted to medium. Higher thinking can use more output tokens, which cost $7.50 per million.
- Check old code. VentureBeat said Google’s migration guide rejects the thinking budget parameter and the minimal setting.
- Use Batch when speed is not needed. It costs half the standard rate.
- Set a spend limit. The site’s guide to a Gemini API spend cap explains how.
- Compare with Carbon. The site’s report on Gemini 4 Carbon covers the larger model.
What is not known about Gemini 4 Hydrogen
- How it performs on tasks outside Google’s chosen benchmarks.
- Rate limits and free-tier terms for developers; Google’s developer post did not state them.
- Whether Flash’s introductory price will be extended past December 31.
- Which honesty score is correct, 94% or 96.7%.
Editor’s take
Editor’s take: Gemini 4 Hydrogen trades a higher price for more speed and measured intelligence than Flash. Independent testing puts it near Argon at the default setting. Whether that is worth double Flash’s introductory price depends on the task. For most users, the model will simply appear in the Gemini app and Search. Developers should run their own cost test first.
This article is for general information, based on Google, VentureBeat, 9to5Google and Artificial Analysis as of November 13, 2026. The Jolt Journal did not run the model. Prices and features may change.
