Gemini 4 Hydrogen, the modelGoogle has not announced but whose billing rows sit in Google Cloud’s public SKU catalogue, would cost $3.00 to run a workload of 1 million input tokens and 200,000 output tokens at its listed text rates of $1.50 and $7.50 per million, by our arithmetic.
On the same workload, at each company’s published standard rates, OpenAI’s GPT-6.1 Terra costs $2.00, so Hydrogen’s listed price is 1.5 times Terra’s. It is 15 times the $0.20 of Anthropic’s Claude Haiku 5.5 and OpenAI’s GPT-6 Luna, and 37.5% of the $8.00 of Claude Opus 5.5, by our arithmetic.
Business Pill · Tokens Explained
An evergreen Business Engineer explainer on tokens, the unit AI reads, writes and bills in. It teaches the general idea only and does not discuss Gemini or Hydrogen.
The key insight: As we read it, Hydrogen’s catalogue price puts it above OpenAI’s GPT-6.1 Terra and below GPT-6.1 Sol and Claude Sonnet 5.5 on a workload of 1 million input and 200,000 output tokens: $3.00 against $2.00 and $4.00, by our arithmetic, and 15 times the $0.20 of Claude Haiku 5.5 and GPT-6 Luna.
Update, 12 November 2026, 18:30 UTC: Google has launched Hydrogen at the rates used below.Googlelaunched Gemini 4 Hydrogen on 12 November, and Google DeepMind chief executive Demis Hassabis posted on X at 18:11 UTC that Google was “Very excited to release Gemini 4 Hydrogen”. Google’s Gemini API pricing page now lists Hydrogen at $1.50 per million input tokens and $7.50 per million output tokens, the rates the $3.00 workload below was built on, so the comparison now holds at Google’s official prices: 1.5 times GPT-6.1 Terra at OpenAI’s standard rates, by our arithmetic. Our launch piece: Gemini 4 Hydrogen Launches at 2x Gemini 3.8 Flash’s Price.
Update, 12 November 2026, 16:15 UTC: a developer says Hydrogen was added to AI Studio; Google has not announced it. A developer posted on X at 15:54 UTC that “Gemini 4 Hydrogen has been added to AI Studio”, with a screenshot of what the post presents as AI Studio code, naming a model “models/hydrogen”, a “gemini-hydrogen-playground” example and a documentation address for gemini-4-hydrogen. When we checked between 16:00 and 16:10 UTC, that documentation address returned Google’s not-found page, the AI Studio script served to us without sign-in did not contain the word Hydrogen, so we could not confirm the screenshot, and Google’s models page, pricing page and release notes (newest entry 11 November) had no Hydrogen entry. Neither had the Google or Google DeepMind blogs, nor the X accounts of Google AI Studio, Google DeepMind and Logan Kilpatrick. The $3.00 workload below still rests on the Cloud catalogue rates of $1.50 input and $7.50 output per million tokens; we will re-run it if Google publishes different prices.
We priced one fixed job across 14 rows: 1 million input tokens and 200,000 output tokens, for example 100 requests of 10,000 input and 2,000 output tokens each. The cost is the input price plus 0.2 times the output price, using standard paid rates for short prompts, with no caching, batch or volume discounts.
Hydrogen’s rates come from the catalogue’s text rows, which list standard input at $1.50 per million tokens and text output at $7.50. We covered the listing in our first piece on the Hydrogen price rows. Google’s Gemini API models and pricing pages and its Vertex AI pricing page had no Hydrogen entry when we checked at 09:55 UTC on 12 November.
Cost of 1 million input tokens plus 200,000 output tokens at each model’s standard list rates for short prompts, no caching or discounts. Gemini 4 Hydrogen at the $1.50 input and $7.50 text output rows in Google Cloud’s SKU catalogue (Google has not announced the model); Gemini 4 Carbon and Argon at their 1 December rates; Gemini 3.8 Flash at its rate through 31 December; DeepSeek at peak rates. Prices from Google, Anthropic, OpenAI, DeepSeek and Z.ai pricing pages, read 12 November 2026. Totals are our arithmetic.
Where Hydrogen Lands
Of the 13 other rows, six cost more than Hydrogen’s $3.00 and seven cost less, by our arithmetic. Above it are OpenAI’s GPT-6 Astra at $20.00, Anthropic’s Claude Fable 5.2 at $16.00, Claude Opus 5.5 at $8.00, Gemini 4 Carbon and Argon at $8.00 from 1 December, and GPT-6.1 Sol and Claude Sonnet 5.5 at $4.00 each.
Below it are Z.ai’s GLM-5.3 at $2.28, DeepSeek V4 Pro at $2.11 at peak rates, GPT-6.1 Terra at $2.00, Gemini 3.8 Flash at $1.50 through 31 December, DeepSeek’s Flash model at $0.54 at peak, and Claude Haiku 5.5 and GPT-6 Luna at $0.20 each.
By our arithmetic, Hydrogen’s listed price is 75% of Sol and Sonnet 5.5, about 19% of Fable 5.2 and 15% of Astra. Against Carbon and Argon it is 75% at their current $2 and $10 rates and 37.5% once they move to $4 and $20 on 1 December.
Gemini 4 Hydrogen’s $3.00 cost for 1 million input and 200,000 output tokens at its Google Cloud catalogue rates, as a multiple of the same workload on GPT-6 Luna and Claude Haiku 5.5 ($0.20 each), DeepSeek Flash at peak rates ($0.54) and Gemini 3.8 Flash at its rate through 31 December ($1.50). Prices from each company’s pricing page, read 12 November 2026; multiples are our arithmetic.
Against Low-Priced Models
Business Insider reported that a Google employee likened the model being tested to “Google’s previous Flash models”, as set out in our piece on the Hydrogen testing report. At its catalogue price, Hydrogen sits above the low-priced models that the large labs list.
Anthropic lists Claude Haiku 5.5 at $0.10 per million input tokens and $0.50 output for prompts up to 100,000 tokens; OpenAI lists GPT-6 Luna at the same $0.10 and $0.50. DeepSeek lists its Flash model at $0.30 and $1.20 at peak hours and says “Off-peak rates are half of the peak rates”.
On our workload that puts Hydrogen at 15 times Haiku 5.5 and Luna, about 5.6 times DeepSeek Flash at peak and about 11 times it off-peak, by our arithmetic. Compared with Z.ai’s GLM-5.3 at $1.40 input and $4.40 output, Hydrogen costs about 1.3 times as much.
Google’s Own Ladder
Google’s pricing page calls Gemini 3.8 Flash “Our most intelligent Flash model” and lists it at $0.75 input and $3.75 output through 31 December, then $1.50 and $7.50 starting 1 January 2027. On our workload that is $1.50 today and $3.00 from January, by our arithmetic: half of Hydrogen’s listed price now, and the same from 1 January.
The page lists Gemini 4 Carbon and Argon at $2.00 input and $10.00 output through 30 November and $4.00 and $20.00 starting 1 December. On our workload that is $4.00 now and $8.00 from December, which is level with Claude Opus 5.5 at $4 input and $20 output, by our arithmetic.
We have not placed Hydrogen on our AI index, which uses only benchmark results a lab publishes; Gemini 4 Carbon led that index at 59 in October. Google has not launched Hydrogen or published results for it, so there is no price-performance point to plot yet.
The Structural Read
As we read the price pages, the listed Hydrogen rate sits in the middle of the market rather than at its low end: of the 13 other rows we priced, six cost more and seven cost less, by our arithmetic.
As we read Google’s own ladder, Hydrogen’s catalogue rates match the January 2027 rates for Gemini 3.8 Flash, and on our workload the gap to Carbon and Argon widens from $1.00 now to $5.00 after 1 December, by our arithmetic.
Google Gemini API pricing page, on Gemini 4 Carbon
“Every token the model processes is counted and billed, including thinking tokens and internal tool system instructions and function declarations when using grounding.”
Three Implications
FOR DEVELOPERS PICKING A DEFAULT MODEL On this workload Hydrogen’s listed price falls between GPT-6.1 Terra at $2.00 and GPT-6.1 Sol and Claude Sonnet 5.5 at $4.00, by our arithmetic.
FOR GOOGLE’S LINEUPGoogle’s pricing page lists Gemini 3.8 Flash at $1.50 input and $7.50 output starting 1 January 2027, the same as Hydrogen’s catalogue rows.
FOR COST-SENSITIVE BUYERS DeepSeek says off-peak rates are half of peak rates; off-peak, its Flash model costs $0.27 on this workload, by our arithmetic.
The guide says: “The new job of finance is to manage the ratio between the two — work produced per token burned.”
As we read it, a per-workload price is where that ratio starts: the same 1 million input and 200,000 output tokens cost from $0.20 to $20.00 across the 14 rows we priced, by our arithmetic. The guide predates the Hydrogen listing and does not discuss it.
What Is Not Established
A catalogue row is a billing entry, not a launch. Google has not announced Hydrogen, its context window, its availability or its benchmark results, and the final public price could differ from the rows we read on 12 November.
Our workload is one illustrative mix. Models use different numbers of tokens for the same task, and Google’s page says output prices include thinking tokens, so cost per finished task can differ from cost per million tokens. Caching, batch, flex, peak and regional rates change every total; open-weight models price tokens differently again.
This Week In The AI Supercycle
The Great Convergence of Capital, Power, and Geopolitics
This week’s issue: for two years the AI build-out was priced as a supply story, and last week as a credit story. This week it was priced as a permission story, and the answer arrived as a cost, not a ban.
At its catalogue rates, Gemini 4 Hydrogen would cost $3.00 for 1 million input and 200,000 output tokens, by our arithmetic: 1.5 times GPT-6.1 Terra, 15 times Claude Haiku 5.5 and GPT-6 Luna, and 37.5% of Claude Opus 5.5 at their published standard rates.
95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. Hydrogen’s rates are the text rows of Google Cloud’s public SKU catalogue as we read them on 12 November 2026; Google has not announced the model. Every other price comes from the company’s own pricing page, read the same day: Google’s Gemini API pricing, Anthropic’s Claude API pricing, OpenAI’s API pricing, DeepSeek’s models and pricing page (peak rates unless stated) and Z.ai’s pricing.
We cross-checked the Anthropic, OpenAI and Gemini 4 rates against our own OpenRouter snapshot. Workload totals, ranks and multiples are our arithmetic. Google was not contacted for this piece.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
