Updated pricinggooglegemininew-models

Gemini 4 Carbon: Google's Most Capable Model Costs the Same as Argon, and Both Double on December 1

Short answer. Gemini 4 Carbon, released on November 10, 2026, costs $2.00 per million input tokens and $10.00 per million output tokens through November 30, 2026. From December 1 it costs $4.00 and $20.00. Those are exactly Gemini 4 Argon’s prices. A request of 10,000 input and 1,000 output tokens costs $0.03, then $0.06.

In plain words. Google’s newest and strongest model costs the same per token as the one before it. Both prices double on December 1. A token is a short piece of text, about three-quarters of a word. Carbon thinks longer on hard tasks, so a task can use more tokens.

What Google announced

The Gemini API release notes carry an entry dated November 10, 2026: “Gemini 4 Carbon ( gemini-4-carbon ) generally available (GA): Released Gemini 4 Carbon ( gemini-4-carbon ), our most capable reasoning model built for sustained agentic coding, multimodal knowledge work, scientific reasoning, and cybersecurity defense, available at introductory pricing of $2.00 per 1M input tokens and $10.00 per 1M output tokens through November 30, 2026.”

So this is a general availability release, not a preview, and it comes with an end date for its first price. Google’s deprecations page, read on November 11, lists the model as “gemini-4-carbon | November 10, 2026 | No shutdown date announced |”. The new-models scan that AI Signal runs every morning found it on Google’s pricing page the next day, and Gemini 4 Carbon has had its own page since November 11.

The prices, line by line

Google’s pricing page prints each paid-tier figure with its date. For Gemini 4 Carbon, input is “$2.00 through November 30, 2026.” and “$4.00 starting December 1, 2026.”; output is “$10.00 through November 30, 2026.” and “$20.00 starting December 1, 2026.”; context caching is “$0.10 through November 30, 2026.” and “$0.20 starting December 1, 2026.”, with “Implicit caching only (explicit caching is not supported).” The output line is labelled “Output price (including thinking tokens)”, and there is no second price for long prompts.

Per million tokens, standard paid tier, read on November 11, 2026, with the closest model of the previous Pro generation for comparison:

ModelInputOutputCached input10,000 in + 1,000 out
Gemini 4 Carbon, through November 30$2.00$10.00$0.10$0.03
Gemini 4 Carbon, from December 1$4.00$20.00$0.20$0.06
Gemini 4 Argon, through November 30$2.00$10.00$0.10$0.03
Gemini 4 Argon, from December 1$4.00$20.00$0.20$0.06
Gemini 3.1 Pro Preview, prompts up to 200K$2.00$12.00$0.20$0.032

The request column has no cache: 10,000 input tokens at $2.00 per million is $0.02 and 1,000 output tokens at $10.00 is $0.01, so $0.03 on Carbon until November 30. At list price it is $0.04 plus $0.02, $0.06. On Gemini 3.1 Pro Preview it is $0.02 plus $0.012, $0.032. Gemini 3.1 Pro Preview prints “$2.00, prompts <= 200k tokens”, “$12.00, prompts <= 200k tokens” and “$0.20, prompts <= 200k tokens”, with higher prices above 200K; the request above stays in the first band. Google describes it as “Our 3rd generation Pro model, built for multimodal understanding, agentic capabilities, and vibe-coding.”

Carbon against Argon and Gemini 3.1 Pro Preview

Each ratio below is the larger price divided by the smaller, with the model that costs more.

ComparisonInputOutputCached input10,000 in + 1,000 out
Carbon vs Argon, through November 301× (same)1× (same)1× (same)1× (same)
Carbon vs Argon, from December 11× (same)1× (same)1× (same)1× (same)
Carbon vs Gemini 3.1 Pro Preview, through November 301× (same)1.2×, 3.1 Pro Preview costs more2×, 3.1 Pro Preview costs more1.1×, 3.1 Pro Preview costs more
Carbon vs Gemini 3.1 Pro Preview, from December 12×, Carbon costs more1.7×, Carbon costs more1× (same)1.9×, Carbon costs more

Until November 30, Google’s strongest model is slightly cheaper per request than Gemini 3.1 Pro Preview: same input, $10.00 against $12.00 for output. From December 1 the order turns. Carbon’s $0.06 request is 1.9 times the $0.032 of Gemini 3.1 Pro Preview, whose prices have no end date on the page.

Against Argon there is nothing to divide. Google’s “What’s new in Gemini 4 Carbon” guide says it plainly: “Gemini 4 Carbon and Gemini 4 Argon share the same 1M unified context window, built-in tools, and per-token pricing.”

Same price per token, not always per task

The same guide adds one line about cost: “Carbon spends more thinking tokens per turn than Argon on complex tasks.” Thinking tokens are billed as output: the pricing page’s note for the model says “Every token the model processes is counted and billed, including thinking tokens and internal tool system instructions and function declarations when using grounding.” So two requests with the same prompt can cost different amounts on the two models, because Carbon may write more output tokens before it answers. Google does not say how many more, and this post does not guess. The table above prices a request of a fixed size; a real task on Carbon can produce more than 1,000 output tokens where Argon produces fewer.

That is also how Google tells the two apart. On its pricing page Carbon is “Our most capable reasoning model, built for complex software engineering, deep mathematical and scientific problem solving, high factual accuracy, autonomous multi-step agents, and long-decode generation up to 1M tokens.” Argon, on the same page, is now “Our efficient Gemini 4 frontier reasoning model. It delivers Gemini 4 intelligence with faster responses and lower token spend per turn compared to Gemini 4 Carbon across complex coding, agentic, math, and science tasks.” When AI Signal last read the page before the launch, on November 6, 2026, the first of those descriptions, with “Our flagship reasoning model” in place of “Our most capable reasoning model” and without “high factual accuracy”, was Argon’s.

Limits and free use

  • No free tier. “Gemini 4 Carbon is not available on the Free tier.” Every free-tier cell of its price table reads “Not available”.
  • Context and output. The model page gives “Context window (unified input and output token limit) 1,048,576” and “Max output 65,536 (default synchronous) / up to 1,048,576 (with long decode )”. Input, thinking and output share that one window.
  • Caching. Only implicit caching, as quoted above. The guide adds: “Implicit caching is enabled by default for prompts of at least 4,096 tokens and discounts cached input tokens by 95% on cache hits.” The 95% is $0.10 against $2.00 now, and $0.20 against $4.00 from December 1.

In the Gemini app, not the API

Google’s subscriptions page for Spain, read with a browser on November 11, 2026, now heads its Google AI Ultra card “Incluye Gemini 4 Carbon” and lists “El mayor acceso a Gemini 4 Carbon” among Ultra’s lines, next to “El mayor acceso a Gemini 4 Argon”. That is the Gemini app sold as a monthly plan, at “€99,99 EUR / mes” or “€219,99 EUR / mes”, not the API. The API prices above come only from Google’s API pages, and this post does not compare the two. The plans are on the Google AI plans page.

What else changes on December 1

December 1, 2026 is a busy day for Google’s price list. Both Gemini 4 models double that day. Two older Gemini models shut down the same day, as covered in the Gemini 2.5 Pro and 2.5 Flash shutdown post. Every current Google price is on the Google API pricing page, and Carbon is on the new models page.

Update, November 13, 2026: Google released Gemini 4 Hydrogen on November 12 at $1.50 per million input tokens and $7.50 per million output tokens, with no end date on its pricing page. The comparison with Carbon and Argon is in the Gemini 4 Hydrogen post.

How we checked

Every figure and sentence above comes from Google’s own pages, read on November 11, 2026 through the Markdown version Google serves for each page: the Gemini API pricing page, the release notes, the models and deprecations pages, the Gemini 4 Carbon and Gemini 4 Argon model pages and the two “What’s new” guides. The Spanish subscriptions page was read the same day with a headless browser, from the “España” view in euros. Our dataset rows for Carbon and Argon match those readings; the increase on December 1 is recorded in advance on the price-changes page as an announced provider increase, for both models. Prices are standard paid-tier list prices in US dollars, without Batch, Flex or Priority. The request costs are our own arithmetic from the per-million prices.

Frequently asked questions

How much does Gemini 4 Carbon cost?

$2.00 per million input tokens, $10.00 per million output tokens and $0.10 for cached input through November 30, 2026, on Google's pricing page read on November 11, 2026, and $4.00, $20.00 and $0.20 from December 1, 2026. A request of 10,000 input and 1,000 output tokens costs $0.03, and $0.06 from December 1.

Is Gemini 4 Carbon more expensive than Argon?

Not per token: Google's pricing page prints the same figures for both, and its What's new guide says the two 'share the same 1M unified context window, built-in tools, and per-token pricing.' The same guide says 'Carbon spends more thinking tokens per turn than Argon on complex tasks.' Google does not say how many more.

When does the Gemini 4 Carbon price go up?

On December 1, 2026: the pricing page prints '$2.00 through November 30, 2026' and '$4.00 starting December 1, 2026' for input, '$10.00 through November 30, 2026' and '$20.00 starting December 1, 2026' for output, and Gemini 4 Argon doubles on the same day.

Is Gemini 4 Carbon free?

No: Google's pricing page says 'Gemini 4 Carbon is not available on the Free tier' and prints 'Not available' in the free tier column for every price, so it is sold only on the paid tier of the Gemini API.

Drafted with AI assistance from the cited source and reviewed by Álvaro Lucero Rufino, who is responsible for what stays published here. Automated publishing was discontinued in September 2026; posts that no longer met that standard were removed rather than left up. This is original explanatory writing, not a reproduction — read the source above for the full story.