Dev.to AI 🤖 Ai 👁 0 📖 7 min read

Gemini 4 Neon Pricing: $3 Input, $15 Output and What a Call Really Costs

Gemini 4 Neon Pricing: Complete Rate Card, Real-World Costs, and Thinking-Level Tradeoffs Gemini 4 Neon pricing is $3 per 1M input tokens and $15 per 1M output tokens on the Gemini API's Standard paid tier, flat across

Gemini 4 Neon Pricing: $3 Input, $15 Output and What a Call Really Costs

Gemini 4 Neon Pricing: Complete Rate Card, Real-World Costs, and Thinking-Level Tradeoffs

Gemini 4 Neon pricing is $3 per 1M input tokens and $15 per 1M output tokens on the Gemini API's Standard paid tier, flat across its 1M-token context window. Thinking tokens bill at the output rate, cached input costs $0.15 per 1M, Batch and Flex halve every rate, and Priority costs $5.40 and $27. There's no free API tier.

Try Apidog today

Google released Neon on November 18, 2026, as the middle model of the Gemini 4 family, between the fast Hydrogen and the heavyweight Carbon. Below: the rate card, what real calls cost, how thinking levels move your bill, and how Neon compares with its siblings before and after December 1.

How much does Gemini 4 Neon cost per 1M tokens?

Gemini 4 Neon (model ID gemini-4-neon) has four paid tiers on Google's pricing page. All prices are in USD per 1M tokens:

Tier Input Output incl. thinking Cached input
Standard $3.00 $15.00 $0.15
Batch $1.50 $7.50 $0.075
Flex $1.50 $7.50 $0.075
Priority $5.40 $27.00 $0.27
Free tier Not available Not available Not available

Three details matter most:

  1. Thinking is output. The output column reads "including thinking tokens", so every token Neon spends reasoning costs $15 per 1M on Standard, even though you never see it in the answer.
  2. Caching is implicit only. You get the $0.15 rate when Google's cache matches a repeated prompt prefix; you can't create a cache object yourself.
  3. No free tier. Google's pricing page lists Neon as not available on the Free tier. Hydrogen is the Gemini 4 model you can call for $0 (see our Gemini 4 Hydrogen free guide).

Gemini 4 Neon pricing tiers per 1M tokens: Batch and Flex at $1.50 input and $7.50 output, Standard at $3 input and $15 output, Priority at $5.40 input and $27 output, cached input at $0.15 on Standard, and no free API tier

What do the four Gemini 4 Neon tiers mean?

The four tiers sell the same model at different speed guarantees. Batch returns results asynchronously at half price, Flex is half price for traffic that tolerates slower processing, and Priority costs 80% more than Standard to stay fast when capacity is tight. Pick by deadline:

Your workload Tier Why
Interactive coding assistant, chat Standard A user is waiting; $3/$15 is the baseline
Nightly code review, eval runs, document backfills Batch Nobody waits; you pay $1.50/$7.50
Background agents that can retry later Flex Same half price, without packaging a batch file
Customer-facing agent with an uptime promise Priority $5.40/$27 buys you capacity under load

Grounding has its own meter. On Standard and Priority, Google Search grounding includes 5,000 free search requests per month, shared across Gemini 3 and newer models, then costs $14 per 1,000. Google Maps grounding follows the same 5,000-free, then $14 per 1,000 pattern. The pricing page lists no grounding price for Batch or Flex.

How do thinking levels change your Gemini 4 Neon bill?

Neon supports four thinking levels: low, medium, high and xhigh. The level doesn't change the price per token; it changes how many output tokens Neon burns before it answers.

Artificial Analysis ran its Intelligence Index at every level and published the output tokens each run used. Multiply by the $15 Standard output rate and you get the output cost of the full run:

Thinking level Intelligence Index Output tokens Output cost at $15/1M Time per task
low 44 28.3M $424.50 3.7 min
medium 55 62.9M $943.50 8.3 min
high 60 123M $1,845 14.8 min
xhigh 61 283M $4,245 21.5 min

Two things jump out. Going from high to xhigh buys one index point for 2.3x the output tokens. And medium delivers most of high's score for about half the tokens. Artificial Analysis puts Neon's headline cost at $6.44 per index task at xhigh.

Gemini 4 Neon output tokens by thinking level on the Artificial Analysis Intelligence Index: low 28.3M tokens for $424.50, medium 62.9M for $943.50, high 123M for $1,845 and xhigh 283M for $4,245 at $15 per 1M output tokens

The practical rule: start your tests at medium, move a task to high only when you can show it fails at medium, and save xhigh for the rare problem where one extra point of quality is worth more than double the cost.

What does one Gemini 4 Neon call cost?

Here are three common request shapes, priced on Standard. Output figures include thinking tokens.

1. A code review call. 20,000 input tokens (the diff plus context), 3,000 output tokens:

  • Input: 20,000 × $3 / 1M = $0.060
  • Output: 3,000 × $15 / 1M = $0.045
  • Total: $0.105 per call, or $0.0525 on Batch

Run 1,000 of those a day for 30 days and you spend $3,150 a month on Standard, or $1,575 if the reviews can wait for Batch.

2. An agent loop with a cached prefix. Ten turns, each sending the same 50,000-token system prompt and tool list, plus 5,000 new input tokens and 2,000 output tokens:

  • With a cache hit on the prefix: 50,000 × $0.15 / 1M + 5,000 × $3 / 1M + 2,000 × $15 / 1M = $0.0525 per turn, $0.53 for the loop
  • With no cache hit: 55,000 × $3 / 1M + 2,000 × $15 / 1M = $0.195 per turn, $1.95 for the loop

That's a 73% difference from caching alone. Because Neon's caching is implicit, you can't force the hit; you raise its odds by keeping the stable part of the prompt at the start and byte-for-byte identical across turns.

3. A long-document question. 800,000 input tokens (a large codebase or contract set) and 4,000 output tokens:

  • Input: 800,000 × $3 / 1M = $2.40
  • Output: 4,000 × $15 / 1M = $0.06
  • Total: $2.46, or $1.23 on Batch

Neon has no long-context surcharge on the pricing page, so token 900,000 costs the same as token 1. The context window is 1,048,576 tokens shared between input and output, and a single synchronous request returns up to 65,536 output tokens by default.

How does Neon pricing compare with Hydrogen and Carbon?

Neon costs exactly twice Hydrogen per token. The Carbon comparison flips on December 1, when Carbon's introductory rate ends:

Model Input / 1M Output / 1M Code review call (20K in, 3K out) Free tier
Gemini 4 Hydrogen $1.50 $7.50 $0.0525 Yes
Gemini 4 Neon $3.00 $15.00 $0.105 No
Gemini 4 Carbon, through Nov 30 $2.00 $10.00 $0.070 No
Gemini 4 Carbon, from Dec 1 $4.00 $20.00 $0.140 No

For the next eleven days, Carbon is cheaper per token than Neon. If you build on Carbon's $2/$10 today, budget for $4/$20 in December. Our Gemini 4 Carbon pricing breakdown covers that change in detail.

From December 1, Neon sits in the middle: 25% under Carbon, 2x Hydrogen per token. Per task, the gap depends on thinking tokens, so measure cost per request on your own prompts instead of comparing rate cards.

Route by difficulty: Hydrogen for high-volume everyday calls, Neon for complex coding and analysis you run often, and Carbon for the hardest single problems.

How can you cut your Gemini 4 Neon bill?

Five levers, in the order they usually pay off:

  1. Lower the thinking level. medium uses about half the output tokens of high on the Artificial Analysis runs.
  2. Send waiting work to Batch or Flex. Anything nobody reads in real time costs half.
  3. Shape prompts for implicit caching. Keep system prompts and tool definitions at the top and unchanged, so repeated prefixes bill at $0.15 instead of $3.
  4. Cap output. Set a maximum output token count that fits the task.
  5. Route easy calls to Hydrogen. Keep Neon for work that needs it.

Check the Gemini API rate limits page for your tier before you plan throughput.

Measure your real Gemini 4 Neon cost with Apidog CLI

Every response from Google's Interactions API carries a usage object with input, cached, thought and output token counts. That block, times the rate card above, is your real cost per request. Running the same prompts at each thinking level and comparing them is a repetitive job, so hand it to your coding agent.

Apidog CLI gives Claude Code, Codex, Cursor or OpenCode a command-line way to create and run API tests. Set it up once:

npm install -g apidog-cli@latest
apidog login --with-token <YOUR_TOKEN>
npx -y skills add https://apidog.com   # pick your agent when prompted

Then give your agent one prompt:

In Apidog project <id>, create an environment for the Gemini API with
GEMINI_API_KEY and THINKING_LEVEL variables. Add a test case for
POST /v1beta/interactions with model gemini-4-neon and
generation_config.thinking_level set from THINKING_LEVEL. Send our code review
prompt, assert HTTP 200 and non-empty output, and save usage.total_input_tokens,
total_cached_tokens, total_thought_tokens and total_output_tokens. Run it at
low, medium and high, then compute the Standard cost of each run at $3 input,
$0.15 cached and $15 per 1M for thought and output tokens.

The agent creates the environment, checks its test case payload against the schema, then runs it once per level:

apidog environment create gemini-api --project <id> --base-url https://generativelanguage.googleapis.com
apidog cli-schema get test-case-create
apidog cli-schema validate test-case-create --file case.json
apidog test-case create --project <id> --file case.json
apidog run --test-case <caseId> --project <id> -e <envId> --env-var "THINKING_LEVEL=low" -r cli,json
apidog run --test-case <caseId> --project <id> -e <envId> --env-var "THINKING_LEVEL=medium" -r cli,json
apidog run --test-case <caseId> --project <id> -e <envId> --env-var "THINKING_LEVEL=high" -r cli,json

The JSON reports give your agent the token counts to price, so you get a cost-per-level table for your own prompts. Add apidog run ... -r junit to CI with the Apidog CLI CI/CD guide to catch a prompt change that doubles your thinking tokens, or start from Apidog CLI in Claude Code. Apidog's Free plan covers up to 4 users.

FAQ

Is Gemini 4 Neon free? Not through the API. Google's pricing page lists Neon as not available on the Free tier, so you need a paid Gemini API project. Hydrogen is the Gemini 4 model with a free tier.

Are thinking tokens billed separately? No. They bill as output tokens, at $15 per 1M on Standard. The Interactions API's usage object reports them as total_thought_tokens, so you can see how much of each bill is reasoning.

Does Gemini 4 Neon cost more past 200K tokens? No. The pricing page shows one input rate and one output rate for Neon, with no long-context tier, across the full 1,048,576-token window.

How much cheaper is Batch? Exactly half: $1.50 per 1M input and $7.50 per 1M output, with cached input at $0.075. Flex has the same rates.

Is Neon cheaper than Carbon? Not until December 1. Carbon's introductory $2/$10 runs through November 30, then rises to $4/$20, which puts Neon's $3/$15 about 25% below Carbon per token.

📰 Read the original article on Dev.to AI

Originally published by Dev.to AI. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.