Gemini 4 Carbon Pricing: Cost Before and After Dec 1

Gemini 4 Carbon costs $2 per million input tokens and $10 per million output tokens through November 30, 2026. On December 1, 2026 the price doubles to $4 input and $20 output. The Batch tier costs half of whatever the standard price is that day: $1 / $5 now, $2 / $10 from December 1. Output prices include thinking tokens, and the model has no free tier on the Gemini API. These are Google's own numbers from the Gemini API pricing page, checked Nov 11, 2026.

Whatever Gemini 4 Carbon costs you this month, the same traffic costs twice as much on December 1. Below are the full price table, monthly costs for three common workloads, the same workloads on Claude Fable 5.2 and Claude Sonnet 5.5, and what we would do in the next three weeks.

How much does Gemini 4 Carbon cost per million tokens?

All prices are in USD per 1M tokens, paid tier, from Google's Gemini API pricing page (checked Nov 11, 2026).

Other details from the same page that change real bills:

• Google lists no separate rate for prompts above 200K tokens on Gemini 4 Carbon, so a long prompt costs the same per token as a short one.
• The context window is 1,048,576 tokens, shared between input and output. Max output is 65,536 tokens by default, or up to 1,048,576 with long decode (model page, checked Nov 11, 2026).
• Caching is implicit only. Google doesn't support explicit context caching on this model, so repeated prefixes get the cache rate only when its implicit cache hits.
• Flex costs the same as Batch. Priority costs $3.60 / $18 now and $7.20 / $36 from December 1.
• Grounding with Google Search includes 5,000 free requests per month, shared across Gemini 3 and newer models, then $14 per 1,000 requests.
• There is no free tier. Google bills every token, thinking tokens included.

What will Gemini 4 Carbon cost per month?

We priced three workloads we see often, using only the list prices above. Every example assumes no cache hits, so caching would bring each number down.

Google bills thinking tokens as output. Our output numbers are the total output tokens per request, thinking included. If your prompts trigger long reasoning, measure real output tokens in your logs before you trust any estimate, ours included.

Workload 1: support chatbot

• 1,000,000 requests per month
• 1,500 input tokens per request (system prompt, history, user message)
• 300 output tokens per request
• Total: 1,500M input tokens and 300M output tokens

A live chatbot can't wait for Batch results, so the Batch line is there for comparison only. For this workload, the real December price is $12,000 a month.

Workload 2: RAG / document Q&A

• 100,000 questions per month
• 8,000 input tokens per question (retrieved chunks plus the question)
• 500 output tokens per answer
• Total: 800M input tokens and 50M output tokens

RAG is input-heavy, so the input price drives the bill. Trimming retrieved chunks from 8,000 to 5,000 tokens cuts the December input cost from $3,200 to $2,000.

Workload 3: overnight batch summarization

• 200,000 documents per month
• 20,000 input tokens per document
• 1,000 output tokens per summary
• Total: 4,000M input tokens and 200M output tokens

If the summaries only need to be ready by morning, moving to Batch keeps the December bill at the November standard price.

All three workloads at a glance

Gemini 4 Carbon vs Claude Fable 5.2 vs Claude Sonnet 5.5 on price

Same three workloads, same token counts, same no-cache assumption. Claude prices come from Anthropic's pricing page, checked Nov 11, 2026.

Anthropic's page says Claude 4.6 and later models (except Haiku 5.5) bill the full 1M context window at standard rates, so neither Claude model has a long-prompt surcharge either.

Monthly cost, standard tier:

Monthly cost, Batch tier:

On list price, Gemini 4 Carbon matches Claude Sonnet 5.5 until November 30. From December 1 it costs twice as much as Sonnet 5.5 and half as much as Fable 5.2.

A cheaper model that needs two attempts at a task can cost more than a pricier one that gets it right first time. The one sourced quality number we have is the Artificial Analysis Intelligence Index, checked Nov 11, 2026:

Each score applies only to the reasoning setting in brackets. We didn't include a Fable 5.2 row because we only list scores that match the exact model ID. For Fable against Opus, our Fable 5.2 vs Opus 5.5 calculator runs your own token counts, and the Claude Fable 5.2 pricing page has the full Fable rate card. To compare prices across many more models, looot.ai/models lists them side by side.

What to do before December 1

1. Run your evals now. Every test token costs half what it will in December. If you are deciding between Gemini 4 Carbon, Sonnet 5.5 and Fable 5.2, finish the comparison on your own prompts this month.
2. Measure real output tokens. Thinking tokens bill as output. Log a week of real requests and replace our assumptions with your numbers before you budget.
3. Move offline jobs to Batch. Nightly summaries, backfills, classification and eval runs rarely need answers in seconds. On Batch, December costs the same as November standard.
4. Put shared prefixes first. Gemini 4 Carbon only has implicit caching. Keeping system prompts and shared documents at the start of every request gives the implicit cache the best chance to hit at $0.20 instead of $4.
5. Route by task. If Sonnet 5.5 handles your easy requests at half the December Gemini price, send those to Sonnet and keep Gemini 4 Carbon or Fable 5.2 for the hard ones.
6. Check your usage dashboard on December 2 and compare the daily spend with November's.

Testing on a budget? Gemini 4 Carbon has no free tier, but our free AI models page lists Gemini models and APIs you can try at no cost.

FAQ

How much does Gemini 4 Carbon cost?

$2 per million input tokens and $10 per million output tokens through November 30, 2026. From December 1, 2026 it costs $4 input and $20 output. Batch is half the standard price on either date.

When does the Gemini 4 Carbon price increase?

On December 1, 2026. Google's pricing page lists the $2 / $10 rate "through November 30, 2026" and $4 / $20 "starting December 1, 2026".

Does Gemini 4 Carbon have a free tier?

No. Google marks the free tier "Not available" for Gemini 4 Carbon and bills every token, including thinking tokens.

How much is Gemini 4 Carbon on the Batch API?

$1 input and $5 output per million tokens through November 30, 2026, then $2 and $10 from December 1. Flex is the same price as Batch.

Does Gemini 4 Carbon charge more for prompts over 200K tokens?

Google lists no long-context tier for Gemini 4 Carbon. The same per-token price applies across the 1,048,576-token context window.

What is the Gemini 4 Carbon context window?

1,048,576 tokens, shared between input and output. Default max output is 65,536 tokens, or up to 1,048,576 with long decode.

Is Gemini 4 Carbon cheaper than Claude?

It depends on the Claude model. Through November 30 it matches Claude Sonnet 5.5 at $2 / $10. From December 1 it costs twice Sonnet 5.5 and half of Claude Fable 5.2, which is $8 / $40.

Do thinking tokens count as output tokens?

Yes. Google's output price for Gemini 4 Carbon is labeled "including thinking tokens", so reasoning-heavy prompts cost more than the visible answer suggests.

Prices checked Nov 11, 2026 on Google's Gemini API pricing page and Anthropic's pricing page. We will update this post after the December 1 price change.