That is about 9% more than Claude Opus 5.5, which scores 59 at $10.04 a task on the same leaderboard we read at 18:37 UTC on 15 November. Today Carbon costs about 45% less than Opus 5.5 per task, by our arithmetic.
Business Pill · Usage-Based Pricing Explained
An evergreen Business Engineer explainer on usage-based pricing, paying for work done rather than for seats, the model behind per-token API rates and per-task costs. It teaches the general idea only and does not discuss Gemini 4 Carbon or its pricing.
The key insight: As we read it, Gemini 4 Carbon’s cost advantage over Claude Opus 5.5 is a November price: at $5.48 a task it costs about 45% less than Opus 5.5 on the Artificial Analysis leaderboard, and at the prices Google lists from 1 December the same task comes to $10.96, about 9% more, by our arithmetic.
For Carbon’s standard paid tier, the page lists input at “$2.00 through November 30, 2026. $4.00 starting December 1, 2026.” Output, including thinking tokens, is “$10.00 through November 30, 2026. $20.00 starting December 1, 2026.” The context caching price moves from $0.10 to $0.20 per million tokens.
The Batch and Flex tabs move from $1 and $5 to $2 and $10 per million input and output tokens, and Priority from $3.60 and $18 to $7.20 and $36. Gemini 4 Argon carries the same figures on every tab. Gemini 4 Hydrogen, at $1.50 and $7.50, shows no dated change.
Google’s Gemini API changelog entry for 10 November says Carbon was released “available at introductory pricing of $2.00 per 1M input tokens and $10.00 per 1M output tokens through November 30, 2026”. By our arithmetic the introductory window runs 21 days, and 16 days remain from 15 November.
Cost per Artificial Analysis Intelligence Index task for Gemini 4 Carbon and Gemini 4 Argon at current prices and at the prices Google lists from 1 December (our arithmetic: today’s cost doubled at the same token use), against Claude Opus 5.5 and GPT-6 Astra, from the Artificial Analysis leaderboard read at 18:37 UTC on 15 November 2026.
What Was Posted on X
The #2 model on the leaderboard doubles in price on December 1. It’s sitting in plain sight on Google’s pricing page.
Gemini 4 Carbon: Input: “$2.00 through November 30, 2026. $4.00 starting December 1, 2026.” Output: “$10.00 through November 30, 2026. $20.00 starting December 1, 2026.”
$4 in, $20 out. That is exactly what Opus 5.5 costs today.
Right now Carbon is the bargain: 60 on the Artificial Analysis index at $5.48 a task, against Opus 5.5’s 59 at $10.04. At the same token use, December Carbon is about $11 a task. The gap is gone.
If your agent’s unit economics were built on Carbon in November, they expire in 16 days. Run the numbers this week, not when the invoice arrives.
— Kerem — road to $100k (@mkeremturhan) Nov 15 16:04
Kerem Turhan posted on 15 November at 16:04 UTC that the second model on the leaderboard doubles in price on 1 December, quoting Gemini 4 Carbon’s input and output lines from Google’s pricing page. The post compares Carbon’s 60 at $5.48 a task on the Artificial Analysis index with Opus 5.5’s 59 at $10.04 and puts December Carbon at about $11 a task. Those figures match the pages we read; the $10.96 in this piece is our own arithmetic on the leaderboard’s $5.48.
How Vertex AI Bills the Introductory Price
Google Cloud’s Vertex AI pricing page lists the same $2.00 and $10.00 Global rates for Carbon through 30 November and $4.00 and $20.00 from 1 December, with non-global rates of $4.40 and $22.00 from that date. Its footnote says “Promotional pricing provided through 50% credits back on net spend on select models within a given period”.
The same footnote says the promotion does not apply to Gemini 4 Carbon Cyber Permissive, which the page already lists at the $4.00 and $20.00 rates.
Carbon Against Opus 5.5 per Task
Artificial Analysis says each evaluation’s cost is “calculated from input, cache hit, cache write, reasoning, and answer token prices”. Because Google doubles each of Carbon’s token prices, the cost per task doubles at the same token use, from $5.48 to $10.96, by our arithmetic. That is $0.92 a task above Opus 5.5.
On the same leaderboard GPT-6 Astra scores 57 at $6.24 a task. Carbon costs about 12% less than Astra today and about 1.8 times as much at December prices, by our arithmetic. Gemini 4 Argon, at 50 and $4.97 today, comes to $9.94 at December prices, about 1% below Opus 5.5.
Per token, the gap closes rather than reverses. Anthropic’s Claude pricing page lists Opus 5.5 at $4 per million input tokens and $20 per million output tokens, the rates Carbon takes on 1 December. A workload of 1 million input tokens and 200,000 output tokens, with no caching, costs $4.00 on Carbon today and $8.00 from 1 December, the same $8.00 as on Opus 5.5, by our arithmetic. Claude Fable 5.2, at $8 and $40, comes to $16.00.
As we read the pages together, Google lists $4 and $20 per million tokens as Carbon’s rate from 1 December on both its Gemini API and Vertex AI pricing pages, and Vertex describes the current $2 and $10 rate as promotional pricing provided through credits back on net spend.
Google, Gemini API pricing page, Gemini 4 Carbon output price, read 15 November 2026
“$10.00 through November 30, 2026. $20.00 starting December 1, 2026.”
Three Implications
FOR TEAMS RUNNING AGENTS ON CARBON At the same token use, a task Artificial Analysis prices at $5.48 today comes to $10.96 from 1 December, by our arithmetic.
FOR TEAMS COMPARING CARBON AND OPUS 5.5 Per token the two match from 1 December at $4 and $20 per million; per index task Carbon is $0.92 above Opus 5.5, by our arithmetic.
FOR BATCH WORKLOADSGoogle’s Batch tab moves from $1 and $5 to $2 and $10 per million input and output tokens on the same date.
The guide says a token “is the unit of intelligence consumed; an agent is a process that consumes tokens to produce work. The new job of finance is to manage the ratio between the two — work produced per token burned.”
As we read it, Google’s 1 December change moves only the price side of that ratio: Carbon’s work per task on the Artificial Analysis index stays at a score of 60 while its cost per task doubles at the same token use, by our arithmetic. The guide predates Gemini 4 and does not discuss it.
Gemini 4 Carbon paid-tier prices per million tokens on Google’s Gemini API pricing page, through 30 November and from 1 December 2026, by tier, as read at 18:34 UTC on 15 November 2026.
What Is Not Established
The $10.96 figure assumes Carbon uses the same tokens per task after 1 December. Artificial Analysis may re-run or re-price its evaluations, and its view per effort setting gives different costs from its one-row-per-model view. The pages we read do not say why Google set a 21-day introductory window.
Cost per task depends on the workload. Caching, batch discounts and enterprise terms change what a customer pays. We did not test the models, and Google, Anthropic and Artificial Analysis were not contacted for this piece.
This Week In The AI Supercycle
AI Lost Its Central Banker, the Lab Oligopoly, and the Future of Open Source
This week’s issue: the Nvidia leadership change, read as the AI build-out losing its central banker: who now finances the GPU landlords, backstops the data centres and sponsors open models, and why two labs are becoming the bank.
Google’s pricing page doubles Gemini 4 Carbon’s token prices on 1 December, to $4 and $20 per million. On the Artificial Analysis leaderboard, where Carbon is second at $5.48 a task, that is $10.96 a task at the same token use, about 9% above Opus 5.5’s $10.04, by our arithmetic.
95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. Gemini 4 Carbon and Argon prices and the dated change are from Google’s Gemini API pricing page as we read it at 18:34 UTC on 15 November 2026; Vertex AI rates and the promotional-credit footnote are from Google Cloud’s Vertex AI pricing page read at 18:40 UTC; the introductory-pricing wording is from the Gemini API changelog entry of 10 November. Index scores and cost per task are from the Artificial Analysis leaderboard (one row per model) read at 18:37 UTC, and its cost method from its Gemini 4 Carbon page.
Claude prices are from Anthropic’s pricing page read at 18:40 UTC. The change was flagged in a post on X by Kerem Turhan at 16:04 UTC. Costs at December prices, ratios and day counts are our arithmetic. We did not test the models, and Google, Anthropic and Artificial Analysis 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.
