Google’s Gemini 4 Hydrogen costs $1.87 per task at its default high setting on the live Artificial Analysis leaderboard, against $7.77 for Gemini 3.8 Flash (high), when we read it at 15:43 UTC on 13 November. Hydrogen scores 51 on the Intelligence Index and Flash scores 31.
By our arithmetic, Hydrogen costs about 24% of Flash per index task, or 76% less, even though its list rates of $1.50 and $7.50 per million tokens are twice Flash’s current $0.75 and $3.75, per Google’s pricing page. It also scores 1 point above Gemini 4 Argon (high) at about 38% of Argon’s $4.97 per task.
Business Pill · Tokens Explained
An evergreen Business Engineer explainer on tokens, the unit AI models are priced in. It teaches the general idea only and does not discuss the leaderboard or these models.
The key insight: As we read it, per-token price and per-task cost now point in opposite directions for Google’s cheaper models: Gemini 4 Hydrogen lists at twice Gemini 3.8 Flash’s rates, yet on the Artificial Analysis leaderboard it costs $1.87 per index task against Flash’s $7.77, about 76% less by our arithmetic, and scores 51 against 31.
The leaderboard lists four Hydrogen settings: xhigh at 54 for $3.16 per task, high at 51 for $1.87, medium at 41 for $0.80 and low at 27 for $0.29. Google’s launch post said 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)”, and the live leaderboard shows the same scores.
Above it, Gemini 4 Carbon (high) scores 60 at $5.48 per task, about 2.9 times Hydrogen (high)’s cost, by our arithmetic. Hydrogen (xhigh) ties Carbon (low) at 54 for about 61% of Carbon (low)’s $5.22. Our guide to the Gemini 4 models sets out each model’s list prices.
List price per million output tokens on Google’s Gemini API pricing page (Gemini 3.8 Flash at its rate through 31 December 2026) and weighted average cost per Intelligence Index task on the Artificial Analysis leaderboard, read at 15:43 UTC on 13 November 2026, for Gemini 3.8 Flash (high) and Gemini 4 Hydrogen (high). The 2x and 24% ratios are our arithmetic.
Why Per-Token and Per-Task Prices Differ
Artificial Analysis says cost per task is calculated from “input, cache hit, cache write, reasoning, and answer token prices”, divided by task count and weighted by each evaluation’s weight in the index. A model that uses fewer tokens per task can cost less per task at higher rates.
Our launch piece reported Hydrogen at twice Flash’s per-token price. On the leaderboard the order flips: Flash (high) costs $7.77 a task and Hydrogen (high) $1.87. Hydrogen (high) also takes 7.9 minutes per task against 14.7 for Flash (high), about 1.9 times faster by our arithmetic; Google’s post says 2.5 times faster.
Against OpenAI and Anthropic at the Same Score
Other labs’ models reach the same scores for less. GPT-6.1 Terra (max) also scores 51, at $0.96 a task, so Hydrogen (high) costs about twice as much; GPT-6.1 Sol (high) scores 50 at $0.59, about a third of Hydrogen’s cost, by our arithmetic.
Claude Opus 5.5 (high) scores 52 at $3.46 a task. At the top, Claude Fable 5.2 (max) leads at 65 for $26.54, as it did in our 11 November read of the index.
The Structural Read
As we read the leaderboard, the list price per token says little on its own about what a task costs: Hydrogen (high) costs about 38% of Argon (high)’s $4.97 per task for 1 more point, by our arithmetic, though the two share a family and Argon’s token rates are 33% higher.
“Weighted average cost per Intelligence Index task. 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.”
Three Implications
FOR TEAMS ON 3.8 FLASH Hydrogen (high) scores 51 at $1.87 per index task against Flash (high)’s 31 at $7.77, the leaderboard shows.
FOR TEAMS ON ARGON Hydrogen (high) scores 1 point more than Argon (high) at about 38% of its $4.97 per task, by our arithmetic.
FOR BUYERS COMPARING LABS GPT-6.1 Terra (max) also scores 51, at $0.96 per task, about half of Hydrogen’s $1.87, by our arithmetic.
The framework says: “The new job of finance is to manage the ratio between the two — work produced per token burned.”
As we read it, the leaderboard measures that ratio directly: Hydrogen (high) produces a 51 for $1.87 per task while Flash (high) produces a 31 for $7.77. The framework predates these models and does not discuss them.
Intelligence Index score against weighted average cost per index task (log scale) for selected current settings of Gemini 4 Hydrogen, Argon and Carbon, Gemini 3.8 Flash and two OpenAI models. Source: Artificial Analysis leaderboard, read 15:43 UTC, 13 November 2026.
What Is Not Established
The leaderboard is live and its figures change. Between our 11 November and 13 November reads, cost per task for Carbon (high) went from $7.76 to $5.48 and for Argon (high) from $9.95 to $4.97, while Flash (high) stayed at $7.77. Cost per task reflects list prices and the tokens each model used on Artificial Analysis’s nine evaluations, not any customer’s bill. We did not test the models.
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.
On the Artificial Analysis leaderboard at 15:43 UTC on 13 November, Gemini 4 Hydrogen (high) scored 51 at $1.87 per task and Gemini 3.8 Flash (high) scored 31 at $7.77. Hydrogen costs twice Flash per token and about 76% less per task, by our arithmetic.
A note on sourcing. Scores, costs per task and times are from the Artificial Analysis leaderboard as we read it at 15:43 UTC on 13 November 2026, and the cost definition from its Gemini 4 Hydrogen model page; the 11 November figures are from our earlier read of the same leaderboard. Per-token rates are from Google’s Gemini API pricing page and the quoted claims from Google’s Hydrogen launch post of 12 November.
Ratios and differences are our arithmetic. We did not test the models, and Artificial Analysis and Google were 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.
