Gemini 4 Hydrogen: the price tag doubles, the bill melts. $1.50 and $7.50 per million tokens, twice Gemini 3.8 Flash today; $1.87 per task against $6.26 at Artificial Analysis

In brief

• Hydrogen: $1.50 per million tokens read, $7.50 per million tokens written, the same price whatever the length of the request, up to 1 million tokens.
• That is 2 times the current price of Gemini 3.8 Flash ($0.75 and $3.75 until 31/12/2026) and 25% less than Gemini 4 Carbon ($2 and $10 until 30/11/2026, then $4 and $20).
• Google is rolling it out "for all users" of Search, worldwide, and in the Gemini app; France is on the country lists for the API, the app and Search's AI Mode.
• Per token, twice the price; per task, no: on its index, Artificial Analysis measures $1.87 per task for Hydrogen at its default setting, against $6.26 for 3.8 Flash at its own, and 6.0 minutes against 12.9. Pushed to xhigh, Hydrogen writes 304 million tokens there, more than four times the median.

The bill

A token is the small piece of text the model counts: "about 4 characters", says Google's documentation. The API is the counter where a program hands its requests to the model, and where every token is billed: the ones the model reads (input) and the ones it writes (output), thinking included. The context is everything it can keep in view at once: here 1 million tokens, shared between what it reads, what it thinks and what it writes.

Here is what Google's price list shows tonight, at the standard paid rate:

Google calls 3.8 Flash's $0.75 and $3.75 an "introductory rate", and $1.50 and $7.50 its "Standard pricing".

Take a typical task: you have the model read a 100,000-token file, about 400,000 characters, and it produces 10,000 output tokens, thinking included. With Hydrogen: $0.15 of reading, $0.075 of writing, so $0.225. A thousand tasks like that: $225. With 3.8 Flash today: $112.50. With Carbon: $300 until 30/11/2026, $600 from 01/12/2026.

So, against Carbon, Hydrogen is 25% cheaper this month; from December, Carbon will cost 2.7 times the price of Hydrogen ($4 versus $1.50, $20 versus $7.50). Against 3.8 Flash, it is 2 times more expensive until 31/12/2026, then the same price. The same task does not necessarily burn the same number of tokens from one model to the next: Google writes that its partners have seen Hydrogen "reducing token usage by 12% to 23%, compared to 3.8 Flash".

Three lines of the price list are worth a look. If you can wait for your answer, batch processing is half price: $0.75 and $3.75 per million tokens. And the free tier exists, with a trade-off written in black and white: on that tier, your data is used to improve Google's products; on the paid tier, it is not. Last, the cache, those tokens already read that the model finds again from one request to the next: $0.075 per million on Hydrogen, a "95% discount", Google writes; $0.075 on 3.8 Flash too until 31/12/2026, then $0.15; $0.10 on Carbon until 30/11/2026, then $0.20. On that line, Hydrogen is not twice the price of 3.8 Flash: it is the same price today, and half the price in January.

As for "flat" pricing, it means a very long request costs no more per token than a short one. On the same page, for requests of more than 200,000 tokens, the reading rate of Gemini 3.1 Pro Preview goes from $2 to $4 per million. Hydrogen has no such step. Neither do Carbon and 3.8 Flash.

What Google promises

Everything below is published by Google itself, in its announcement post, and measured by Google, except the last line: the index is Artificial Analysis's, which Google quotes.

• Speed: compared with 3.8 Flash, Hydrogen is "finishing tasks 2.5 times faster at its high default setting", and the post does not say on which test.
• Made-up claims: "cutting fabricated claims by more than 5x" compared with 3.8 Flash.
• Code: 57% on the Terminal-Bench 4.1 test, versus 18.2% for 3.8 Flash.
• Reasoning: 58.8% on HLE-Diamond, where it "leads its class", a class the post does not define.
• General index: 51 on the Artificial Analysis index at the default setting, versus 50 for Gemini 4 Argon, and 54 at the highest setting, named xhigh.

Google is also launching Orchestrator in Google Antigravity, its development platform: instead of one model from start to finish, it switches between Carbon and Hydrogen during the session. On Terminal-Bench 4.0, Google reports "near-Carbon task success rates while, on average, consuming 24% less quota and completing 19% faster". Quota is the usage limit attached to your subscription in Antigravity.

What Artificial Analysis measures

Figures shown by Artificial Analysis on 12/11/2026 at 19:14 UTC, on its pages for Hydrogen (settings high, the default, and xhigh) and for Gemini 3.8 Flash (settings medium, the default according to Google's documentation, and high). Prices used by Artificial Analysis: $1.50 and $7.50 for Hydrogen, $0.75 and $3.75 for 3.8 Flash, its introductory rate. Artificial Analysis writes: "Evaluations measured independently by Artificial Analysis".

• The bill flips. Twice the price per token, Hydrogen comes to $1.87 per task of this index at its default setting, against $6.26 for 3.8 Flash at its own: 3.3 times less. Even pushed to xhigh: $3.16 against $6.26. That cost, Artificial Analysis writes, is "calculated from input, cache hit, cache write, reasoning, and answer token prices". These are not the tasks of my $0.225 example.
• Speed: 6.0 minutes per task for Hydrogen at its default setting, against 12.9 minutes for 3.8 Flash at its own and 14.8 minutes at the high setting. That is 2.15 times and 2.5 times less time; Google announced 2.5 times.
• Index: 51 at the default setting, 54 at xhigh (16th score out of 78); 3.8 Flash is at 30 and 31; the median of comparable models is 41.
• Talkative at xhigh: 304 million tokens written to run the whole index, against a median of 72.8 million; the page ranks it 76th out of 77 on that criterion. At $7.50 per million, that is $2,280 in writing alone. At the default setting: 110 million.

The price per token does not tell the whole story. This time, what it leaves out works in Hydrogen's favour.

And in France?

If you type into Google. Robby Stein, VP of Product for Google Search, writes that "Hydrogen is now available globally in Google Search for all users", and calls it "the biggest model upgrade to Search yet". What changes, according to him: multi-part questions better understood, more links, more tables and images, "across all supported AI Mode languages". France and French are on the AI Mode list on Google's help page.

If you use the Gemini app. The French subscriptions page names Hydrogen in every plan: free, Google AI Plus at €4.99 a month, Google AI Pro at €21.99, Google AI Ultra from €99.99. Without a subscription, you do not pick it: the app decides on its own. A request that needs more reasoning may go through Hydrogen; most go to Flash-Lite. The menu to select it starts at Google AI Plus.

If you pay for an API. France is on the list of countries for the Gemini API and Google AI Studio, and the changelog lists Hydrogen as generally available since 12/11/2026.

What we don't know yet

• How long the price holds. The price list shows no end date for Hydrogen's $1.50 and $7.50, while it shows one for Carbon and for 3.8 Flash.
• What it gives on your own tasks. The $1.87 and $6.26 are for the tasks of the Artificial Analysis index, not yours.

My verdict

Kira asked me whether Google had just slashed prices. I turned the screen towards her: three rows, two deadlines.

On the price tag, no. Google has put on a brand-new model the price tag the old one was due to take on 1 January, seven weeks early. If you run 3.8 Flash today, moving to Hydrogen doubles your price per token. Your bill is another matter: on the Artificial Analysis index, a task comes to $1.87 with Hydrogen against $6.26 with 3.8 Flash. What is left is to redo the sum on your own tasks. If you were paying Carbon for everyday tasks, then yes, you save: 25% this month, far more in December. Carbon's detailed accounts are in Gemini 4 Carbon's bill, task by task, by Kira.

Cheaper than Carbon, yes. Cheaper per token than yesterday, no. Cheaper per task? At Artificial Analysis, yes.

Words to remember

• Token: the unit the model counts and Google bills; about 4 characters.
• Input, output: the tokens the model reads, and the ones it writes; thinking tokens are billed as output.
• Context: the amount of text the model can handle in a single request.
• API: access for programs, billed by usage, unlike a subscription.
• Batch processing: you send requests in bulk and accept a wait, in exchange for a lower price.
• Thinking level: the dose of reasoning asked of the model; Hydrogen has four, from low to xhigh, with high as the default.

Source sheet

• Google: "Introducing Gemini 4 Hydrogen", by Koray Kavukcuoglu, SVP, Google DeepMind and Chief AI Architect, Google, 12/11/2026; "Google Search gets a major upgrade with Gemini 4 Hydrogen", by Robby Stein, VP of Product, Google Search, 12/11/2026.
• Google documentation: Gemini API price list, What's new in Gemini 4 Hydrogen, What's new in Gemini 3.8 Flash, changelog, model page, available regions, token definition.
• Google consumer pages: subscriptions in France, Gemini app models, Search AI Mode, Gemini app countries.
• Artificial Analysis: Gemini 4 Hydrogen pages, xhigh and high settings; Gemini 3.8 Flash pages, medium and high settings.
• On Silicium Rose: Gemini 4 Carbon: same price per token as Argon, not the same bill, by Kira; our 28/10/2026 article on Gemini 4 Carbon, by White Rose.
• Links: the announcement post · the Search post · the price list · the what's-new guide · subscriptions in France · Artificial Analysis, Hydrogen · Artificial Analysis, 3.8 Flash

Source passages

Pages read on 12/11/2026 by direct request (HTTP 200), including the Artificial Analysis pages S11S15S16S17, re-read at 19:14; the first read of S11, at 18:54, went through Firecrawl. Times in UTC. The ai.google.dev/gemini-api/docs addresses redirect to aistudio.google.com/docs; the text read is the Markdown version of each page.

Show the 32 sources

Ref.	Quote (original language)	URL	Read at
S1↑1↑2↑3↑4↑5↑6+2	Title, date, author: "Introducing Gemini 4 Hydrogen" ; "Nov 12, 2026" ; datePublished "2026-11-12T18:00:00+00:00" ; "Koray Kavukcuoglu" ; "SVP, Google DeepMind and Chief AI Architect, Google"	https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-hydrogen/	18:51
S1↑1↑2↑3↑4↑5↑6+2	"With Hydrogen, we’re also bringing Gemini 4 to everyone — for all tiers of Google AI Studio, the Gemini API, Gemini for business, the Gemini app, Google Search, and paid Google AI Pro and Ultra subscribers in Google Antigravity."	idem	18:51
S1↑1↑2↑3↑4↑5↑6+2	"Hydrogen surpasses Argon on the Artificial Analysis Intelligence Index at its default high thinking level (51 vs. 50, and reaching 54 at its highest xhigh thinking setting), while often completing tasks faster than models in its class. Gemini 4 Hydrogen is priced flat across its entire 1 million-token context window at $1.50 per million input tokens and $7.50 per million output tokens."	idem	18:51
S1↑1↑2↑3↑4↑5↑6+2	"To help make this multi-model collaboration automatic, we’re introducing Orchestrator in Google Antigravity. Rather than using a single model from start to finish, Orchestrator dynamically switches between models like Carbon and Hydrogen within the same session. On Terminal-Bench 4.0, Orchestrator reached near-Carbon task success rates while, on average, consuming 24% less quota and completing 19% faster."	idem	18:51
S1↑1↑2↑3↑4↑5↑6+2	"Beyond multi-model workflows, Hydrogen is a major standalone upgrade for teams migrating from Gemini 3.8 Flash. It delivers significant leaps in both speed and intelligence — finishing tasks 2.5 times faster at its high default setting while more than tripling Terminal-Bench 4.1 scores (57% vs. 18.2%), cutting fabricated claims by more than 5x, and adapting cleanly when a tool call fails or a search returns empty."	idem	18:51
S1↑1↑2↑3↑4↑5↑6+2	"in early testing across complex graphs, SQL, and data-heavy enterprise workflows, our partners have seen Hydrogen run 2x to 2.8x faster while reducing token usage by 12% to 23%, compared to 3.8 Flash."	idem	18:51
S1↑1↑2↑3↑4↑5↑6+2	"For deep domain reasoning and analysis, it leads its class on HLE-Diamond (58.8%), ranks #1 overall on Vals Tax Agent Bench (52.3%), ties for #1 on the Vals Legal Research Bench (55.8%), and leads its class on AutomationBench-AA."	idem	18:51
S1↑1↑2↑3↑4↑5↑6+2	"Hydrogen’s four thinking levels — low, medium, high (default), and a new xhigh tier"	idem	18:51
S1↑1↑2↑3↑4↑5↑6+2	"Starting today, Hydrogen is rolling out globally across our developer, enterprise, and consumer surfaces" ; "For everyone: Start using it in the Gemini app model drop down" (footnote mark 1) "and in Google Search for all users, and in Gemini in Google Sheets starting with Google AI Pro and Ultra subscribers." ; footnote 1: "In the Gemini app, free users will get access to Gemini 4 Hydrogen via auto model selection when it’s the appropriate model for the prompt."	idem	18:51
S2↑1↑2↑3↑4	Title, date, author: "Google Search gets a major upgrade with Gemini 4 Hydrogen" ; "Nov 12, 2026" ; datePublished "2026-11-12T18:00:00+00:00" ; "Robby Stein" ; "VP of Product, Google Search"	https://blog.google/products-and-platforms/products/search/gemini-4-hydrogen-search-upgrade/	18:51
S2↑1↑2↑3↑4	"Today, Google Search is bringing cutting-edge intelligence to everyone with Gemini 4 Hydrogen. This is the biggest model upgrade to Search yet." ; "Hydrogen is now available globally in Google Search for all users, delivering more helpful, visual, and intuitive responses with improved model capabilities"	idem	18:51
S2↑1↑2↑3↑4	"Just ask in your natural flow of thought — including questions with multiple parts, hard-to-understand typos, or highly specific requests." ; "connecting you to a broader range of helpful links than before" ; "You’ll see more images, structured responses, comparison tables, and custom tools built for you." ; "No matter what language you ask in, you’ll now get more highly nuanced and helpful responses across all supported AI Mode languages."	idem	18:51
S3↑1↑2↑3↑4↑5↑6+6	Gemini 4 Hydrogen, Standard, "Paid Tier, per 1M tokens in USD" : "Input price" : "Free of charge" (Free Tier), "$1.50" ; "Output price (including thinking tokens)" : "Free of charge", "$7.50" ; "Context caching price" : "$0.075" ; "Used to improve our products" : "Yes" (Free Tier), "No" (Paid Tier) ; Batch : "$0.75", "$3.75"	https://ai.google.dev/gemini-api/docs/pricing	18:52
S3↑1↑2↑3↑4↑5↑6+6	Gemini 3.8 Flash, Standard : "Input price" : "$0.75 through December 31, 2026. $1.50 starting January 1, 2027." ; "Output price (including thinking tokens)" : "$3.75 through December 31, 2026. $7.50 starting January 1, 2027."	idem	18:52
S3↑1↑2↑3↑4↑5↑6+6	Gemini 4 Carbon, Standard : "Input price" : "Not available" (Free Tier), "$2.00 through November 30, 2026. $4.00 starting December 1, 2026." ; "Output price (including thinking tokens)" : "$10.00 through November 30, 2026. $20.00 starting December 1, 2026." ; "Gemini 4 Carbon is not available on the Free tier."	idem	18:52
S3↑1↑2↑3↑4↑5↑6+6	Gemini 3.1 Pro Preview : "$2.00, prompts <= 200k tokens $4.00, prompts > 200k tokens"	idem	18:52
S3↑1↑2↑3↑4↑5↑6+6	Cache, Standard tab: Hydrogen "Context caching price" "$0.075" ; 3.8 Flash "$0.075 through December 31, 2026. $0.15 starting January 1, 2027." ; Carbon "$0.10 through November 30, 2026. $0.20 starting December 1, 2026." ; S4 : "Cached input tokens: $0.075 per 1M tokens via implicit caching (95% discount)"	idem	18:52, re-read 19:14
S4↑1↑2	"Gemini 4 Hydrogen is available on the free tier of the Gemini API and in Google AI Studio, and uses flat standard pricing on the paid tier across the full 1M context window" ; "Recommended upgrade from Gemini 3.8 Flash and earlier Gemini 3 and 2.5 models." ; "Free tier: Free of charge (rate-limited; see rate limits)"	https://aistudio.google.com/docs/whats-new-gemini-4-hydrogen	18:52
S5↑1↑2	"November 12, 2026" ; "Gemini 4 Hydrogen (gemini-4-hydrogen) generally available (GA)" ; "Available on the free tier and at paid-tier pricing of $1.50 per 1M input tokens and $7.50 per 1M output tokens"	https://ai.google.dev/gemini-api/docs/changelog	18:52
S6↑1↑2	"Context window (unified input and output token limit) 1,048,576" ; "Max output 65,536 (default synchronous) / up to 1,048,576 (with long decode)" ; S4 : "Hydrogen shares one 1,048,576-token window across input, thinking, and output tokens" ; models list: "Our previous-generation Flash model" (Gemini 3.8 Flash)	https://ai.google.dev/gemini-api/docs/models/gemini-4-hydrogen	18:52
S7↑1↑2↑3	"The Gemini API and Google AI Studio are available in the following countries and territories" ; "France"	https://ai.google.dev/gemini-api/docs/available-regions	18:52
S8↑1↑2↑3↑4	"France" ; "Sans frais" : "€0 EUR/mois avec un compte Google", "Includes Gemini 4 Hydrogen" ; "Google AI Plus" : "€4,99 EUR/mois", "Access to Gemini 4 Hydrogen" ; "Google AI Pro" : "€21,99 EUR/mois", "Access to Gemini 4 Hydrogen" ; "Google AI Ultra" : "À partir de : €99,99 EUR / mois", "Highest access to Gemini 4 Hydrogen" ; "Limites de débit de base pour le modèle d’agent dans Google Antigravity, notre plate-forme de développement agentique"	https://gemini.google/fr/subscriptions/?hl=fr	18:53
S9↑1↑2↑3	"No Google AI Plan" : "Auto or Flash-Lite*" ; "Google AI Plus" : "Flash-Lite" "Hydrogen" "Pro" ; "*If auto model selection is turned on, Gemini will choose the appropriate model for each prompt. Most prompts will use Flash-Lite for quick answers. Prompts that require more reasoning may use Hydrogen. If auto model selection is turned off, all responses will use Flash-Lite."	https://support.google.com/gemini/answer/18648722?hl=en	18:53
S10↑1↑2↑3	"AI Mode is available in the following countries, territories, and languages." ; "France" ; "French" ; "Gemini 4 Hydrogen in AI Mode is helpful for fast, all-around assistance on most questions."	https://support.google.com/websearch/answer/16011537?hl=en	18:52
S11↑1↑2↑3	"Gemini 4 Hydrogen (Xhigh) scores 54 on the Artificial Analysis Intelligence Index, placing it well above average among comparable models (median: 41)." ; "When evaluating the Intelligence Index, it generated 304M tokens, which is very verbose in comparison to the median of 72.8M." ; "On average, it costs $3.16 per task to evaluate Gemini 4 Hydrogen (Xhigh) on the Intelligence Index." ; "#16 / 78" ; "#76 / 77" ; "At 13.0 minutes per Intelligence Index task, Gemini 4 Hydrogen (Xhigh) is slower than average compared with the median of 12.9 minutes"	https://artificialanalysis.ai/models/gemini-4-hydrogen	19:14
S12↑1↑2↑3	"For Gemini models, a token is equivalent to about 4 characters. 100 tokens is equal to about 60-80 English words."	https://ai.google.dev/gemini-api/docs/tokens	18:54
S14↑1↑2	"The Gemini web app is currently available in more than 70 languages and over 230 countries and territories." ; "France"	https://support.google.com/gemini/answer/13575153?hl=en	18:52
S15↑1↑2↑3↑4↑5	"Gemini 4 Hydrogen (High) scores 51 on the Artificial Analysis Intelligence Index" ; "it generated 110M tokens" ; "On average, it costs $1.87 per task to evaluate Gemini 4 Hydrogen (High) on the Intelligence Index." ; "At 6.0 minutes per Intelligence Index task" ; "Evaluations measured independently by Artificial Analysis" ; "Each evaluation’s cost is calculated from input, cache hit, cache write, reasoning, and answer token prices, divided by task count, and weighted by its Intelligence Index weight."	https://artificialanalysis.ai/models/gemini-4-hydrogen-high	19:14
S16↑1↑2↑3↑4	"Gemini 3.8 Flash (Medium) scores 30 on the Artificial Analysis Intelligence Index" ; "it generated 83.9M tokens" ; "Pricing for Gemini 3.8 Flash (Medium) is $0.75 per 1M input tokens" ; "and $3.75 per 1M output tokens" ; "On average, it costs $6.26 per task to evaluate Gemini 3.8 Flash (Medium) on the Intelligence Index." ; "At 12.9 minutes per Intelligence Index task"	https://artificialanalysis.ai/models/gemini-3-8-flash-medium	19:14
S17↑1↑2	"Gemini 3.8 Flash (High) scores 31 on the Artificial Analysis Intelligence Index" ; "it generated 126M tokens" ; "On average, it costs $7.77 per task to evaluate Gemini 3.8 Flash (High) on the Intelligence Index." ; "At 14.8 minutes per Intelligence Index task"	https://artificialanalysis.ai/models/gemini-3-8-flash	19:14
S18↑1↑2↑3	Table, "Default thinking level" : Gemini 3.8 Flash "medium" ; "Gemini 3.8 Flash is available at an introductory rate of $0.75/1M input tokens and $3.75/1M output tokens through December 31, 2026. Standard pricing of $1.50/1M input tokens and $7.50/1M output tokens takes effect January 1, 2027."	https://ai.google.dev/gemini-api/docs/whats-new-gemini-3.8	19:14
R1↑1↑2↑3↑4↑5↑6	Our calculations, on the prices in S3: 100,000 tokens read + 10,000 written = $0.15 + $0.075 = $0.225 (Hydrogen); $0.075 + $0.0375 = $0.1125 (3.8 Flash); $0.20 + $0.10 = $0.30 then $0.40 + $0.20 = $0.60 (Carbon); times 1,000. 1.50 / 0.75 = 2; 1 − 1.50 / 2.00 = 25%; 4.00 / 1.50 = 20.00 / 7.50 = 2.67. 304 million tokens × $7.50 per million = $2,280; 304 / 72.8 = 4.18 S11. From 12/11/2026 to 01/01/2027: 50 days, i.e. seven weeks. Artificial Analysis S15S16S17: 6.26 / 1.87 = 3.35; 7.77 / 1.87 = 4.16; 6.26 / 3.16 = 1.98; 12.9 / 6.0 = 2.15; 14.8 / 6.0 = 2.47.	—	—