Gemini 4 Hydrogen Pricing: $7.50 Output, 2x Flash’s Rate

Gemini 4 Hydrogen pricing is $1.50 per million input tokens and $7.50 per million output tokens, flat across a 1 million-token window. Google released the model on November 12 as the successor to Gemini 3.8 Flash. It scores 51 on the Artificial Analysis Intelligence Index versus 31 for Flash, but costs twice Flash’s current introductory rate and about 15 times the cheapest rivals per token.

Key takeaways

• Hydrogen costs $1.50 in and $7.50 out per million tokens, double Flash’s rate through December 31.
• It scores 51 on the Artificial Analysis index at default settings, one point above Gemini 4 Argon.
• Artificial Analysis estimates $1.87 per benchmark task, versus $7.77 for Gemini 3.8 Flash.

What is Gemini 4 Hydrogen?

Gemini 4 Hydrogen is Google’s new fast, mid-priced model for coding, documents and multi-step AI agents. It replaces Gemini 3.8 Flash as Google’s everyday model, drops the “Flash” name, and runs across the Gemini API, AI Studio, Gemini for business, the Gemini app and Google Search, all starting November 12, 2026.

Google DeepMind’s Koray Kavukcuoglu called it “our daily driver for agentic workflows” in the launch post. It sits below the frontier pair, Gemini 4 Carbon and Gemini 4 Argon.

The model offers four thinking levels: low, medium, high (the default) and a new xhigh tier. Google says it takes up to 1 million input tokens and can produce up to 1 million output tokens.

Where Hydrogen shows up first

• Developers: Gemini API and AI Studio, on all tiers; Google Cloud lists gemini-4-hydrogen as generally available.
• Consumers: the Gemini app picker labels it “Everyday help,” per 9to5Google. Free users reach it only through automatic model selection.
• Search: Google calls it the biggest model upgrade to Search yet, live for all users globally.
• Coding: Google Antigravity, for paid Google AI Pro and Ultra subscribers only.

Gemini 4 Hydrogen pricing: how much does it cost?

On the paid Gemini API, Hydrogen costs $1.50 per million input tokens and $7.50 per million output tokens, reasoning tokens included. Cached input costs $0.075 per million. Batch and Flex requests run at half the standard rate, according to VentureBeat’s read of Google’s pricing pages.

The rate is flat across the whole 1 million-token window. There is no long-context surcharge, unlike rivals that raise prices above 100,000 or 200,000 tokens.

VentureBeat also noted that Google lists no introductory promotion for Hydrogen’s standard tier. Google used launch discounts on Argon and Carbon, so this is a deliberate change.

Model	Input ($/1M)	Output ($/1M)	AA Intelligence Index	Est. cost per AA task
Gemini 4 Hydrogen (high)	$1.50	$7.50	51	$1.87
Gemini 4 Hydrogen (xhigh)	$1.50	$7.50	54	$3.16
Gemini 4 Hydrogen (medium)	$1.50	$7.50	41	$0.80
Gemini 3.8 Flash (high, intro price)	$0.75	$3.75	31	$7.77
Claude Haiku 5.5 (max)	$0.10	$0.50	36	$0.41
GPT-6 Luna (max)	$0.10	$0.50	28	$0.15
Meta Muse Ember (max)	$0.10	$0.50	36	$0.14
Kimi K3.1 Pro	$2.00	$10.00	46	$3.68

Is Gemini 4 Hydrogen worth double the Flash price?

For agent and coding work, the numbers say yes. Hydrogen finishes the Artificial Analysis suite for an estimated $1.87 per task at high settings, versus $7.77 for Gemini 3.8 Flash. A higher per-token price buys fewer wasted tokens and fewer retries. For simple chat or classification, it is the wrong tool.

Google reported that Hydrogen completes tasks 2.5 times faster than Flash at its default setting. On Terminal-Bench 4.1, it solved 57% of tasks against 18.2% for Flash, by Google’s count.

Google also said early partners saw Hydrogen run 2x to 2.8x faster with 12% to 23% fewer tokens than Flash. Box’s agent rose from 74% to 79% accuracy on its internal eval, according to a chart CEO Aaron Levie posted, as reported by VentureBeat.

Where the math breaks

The xhigh setting is the trap. Artificial Analysis found it lifted the index from 51 to 54 but raised the estimated per-task cost from $1.87 to $3.16. On its Terminal-Bench 4.0 runs, xhigh scored the same 54% as high.

For high-volume, low-difficulty traffic, Hydrogen loses badly. VentureBeat calculated it costs 15 times as much per token as the lowest standard rates for Claude Haiku 5.5, GPT-6 Luna and Muse Ember.

Is Gemini 4 Hydrogen better than Claude Haiku 5.5 or GPT-6 Luna?

On capability, yes. Even at medium settings, Hydrogen’s index score of 41 beats Haiku 5.5 and Muse Ember at 36 and GPT-6 Luna at 28, all at max reasoning. On price per task, no. Medium Hydrogen costs about $0.80 per Artificial Analysis task, roughly twice Haiku and more than five times Luna or Ember.

The coding gap is the clearest. Artificial Analysis measured Hydrogen at 54% on Terminal-Bench 4.0, against 40.9% for Ember, 32.8% for Haiku and 12.6% for Luna.

Hydrogen still trails the next tier up. Google’s own table shows Claude Sonnet 5.5 at 63.9% on Terminal-Bench 4.1, ahead of Hydrogen’s 57%. Sonnet lists at $2 in and $10 out, about a third more per token.

Open weights remain the budget play. Artificial Analysis puts Xiaomi’s MiMo-V2.6-Flash at about $0.13 per task, but 24 points lower on the index. We covered that price floor in our look at DeepSeek V4.1 Flash 1017.

Who wins and who loses from Hydrogen?

Google wins pricing power. It moved its default workhorse model up a price tier and gave developers a reason to accept it. Gemini 3.8 Flash users lose either way: Flash’s own rate rises to $1.50 and $7.50 on January 1, 2027, matching Hydrogen. Low-cost rivals win the price-sensitive volume Google is walking away from.

Winners

• Alphabet’s API revenue. A doubled list price on its main developer workhorse, with no launch discount, lifts revenue per token if customers stay.
• Anthropic, OpenAI and Meta’s small models. At $0.10 input, Haiku 5.5, Luna and Ember become the obvious home for cost-first workloads.
• Agent builders. Teams running long, multi-step jobs get an Argon-level index score at a fraction of Argon’s $4 and $20 post-introductory list price.

Losers

• Flash-dependent startups. A product priced on Flash’s $0.75 input rate faces double the cost by January, whichever Google model it uses.
• Google’s own cost base. Hydrogen now serves Search for every user worldwide. That is a heavier model on the world’s largest query stream, paid for by Google, not developers.

That last point is the one investors should watch. Alphabet already posted negative free cash flow of $9.1 billion last quarter on AI spending. A pricier model behind Search does not help that line unless ad revenue per query rises with it.

What changes for Gemini 3.8 Flash users?

Flash keeps running for now. Google’s deprecation schedule lists no shutdown date, according to VentureBeat. But its introductory rate of $0.75 and $3.75 ends December 31, 2026. From January 1, 2027, Flash costs the same as Hydrogen, which removes most reasons to stay.

Migration is not a model-name swap. Google’s migration guide, as summarized by VentureBeat, lists these changes:

1. Hydrogen defaults to high reasoning, up from medium on Flash, so bills rise unless you lower it.
2. It rejects the old thinking_budget parameter and the MINIMAL setting.
3. Automatic caching on Google Cloud starts at 4,096 tokens, down from 6,144.
4. Explicit caching is not available at launch.

The default-to-high change is the hidden cost. A team that migrates without touching settings pays for more reasoning on every call. Medium, at an estimated $0.80 per Artificial Analysis task, may be the better match for former Flash traffic.

Should you trust Google’s Hydrogen benchmarks?

Partly. The Artificial Analysis index score of 51 is independent, and it lines up with Google’s claims. Most other headline figures, including the Terminal-Bench 4.1 result, the 5x drop in fabricated claims and the 96.7% agent honesty rate, come from Google’s own tests and have not been replicated.

The Orchestrator claim deserves the most caution. Google said switching between Carbon and Hydrogen in Antigravity reached near-Carbon success on Terminal-Bench 4.0 with 24% less quota and 19% faster completion. Quota is a subscription unit, not dollars, so it does not prove an equal cut in API bills.

Google also said Hydrogen is 2.3x more resilient to indirect prompt injection on the Gray Swan benchmark. That matters for agents that read email and web pages, but it is a relative score, not a guarantee.

What to watch next

Two dated deadlines and two signals will decide whether Hydrogen’s price sticks. The deadlines come from Google’s published Flash pricing; the signals have no date yet.

• December 31, 2026: Gemini 3.8 Flash’s $0.75 and $3.75 introductory pricing ends.
• January 1, 2027: Flash moves to $1.50 and $7.50, the same as Hydrogen. Expect migration volume to peak in December.
• Explicit caching on Google Cloud: its arrival would cut costs for repeated long prompts, the workload Hydrogen’s flat pricing targets.
• Rival price moves: watch whether Anthropic, OpenAI or Meta raise their $0.10 small-model rates or hold them to take Google’s leftover volume.

FAQ

When was Gemini 4 Hydrogen released?

Google released Gemini 4 Hydrogen on November 12, 2026, rolling it out globally across the Gemini API, AI Studio, Gemini for business, the Gemini app and Google Search.

How much does the Gemini 4 Hydrogen API cost?

$1.50 per million input tokens and $7.50 per million output tokens, flat across the 1 million-token context window. Cached input is $0.075 per million, and Batch and Flex are half price.

Is Gemini 4 Hydrogen free?

Free Gemini app users get it only through automatic model selection, which sends most prompts to Flash-Lite. Google AI Plus, at $4.99 a month, lists Hydrogen as its headline model, according to Notebookcheck.

Does Gemini 4 Hydrogen replace Gemini 3.8 Flash?

Yes, as Google’s everyday model. Flash has no announced shutdown date, but its price doubles to match Hydrogen on January 1, 2027.

Is Hydrogen smarter than Gemini 4 Argon?

On the Artificial Analysis Intelligence Index, narrowly: 51 at default high settings versus 50 for Argon, and 54 at xhigh. Google still positions Argon for deeper, long-horizon reasoning.

What thinking levels does Hydrogen support?

Four: low, medium, high (the default) and xhigh. Higher levels raise both quality and token spend.

The bottom line

Gemini 4 Hydrogen is a price increase dressed as a simplification, and a good model anyway. For coding agents and document-heavy workflows, it beats Flash on both quality and cost per finished task. For bulk, low-difficulty traffic, Haiku 5.5, Luna and Ember are far cheaper. Run Hydrogen at medium before paying for high, and skip xhigh unless a test on your own data proves it pays.

Sources

• Google: Introducing Gemini 4 Hydrogen (November 12, 2026)
• Google: Gemini 4 Hydrogen for developers
• Google: Search gets a major upgrade with Gemini 4 Hydrogen
• VentureBeat: Google launches Gemini 4 Hydrogen
• 9to5Google: Gemini 4 Hydrogen rolling out to Gemini app
• Notebookcheck: What free users and Google AI Plus get

Wealth Engine researches and drafts with AI tools and checks every figure against the sources above. How we report.