Gemini API Pricing 2026: Flash, Argon and the 2027 Increase

The Gemini API costs $0.75 per million input tokens and $3.75 per million output tokens on Gemini 3.8 Flash through 31 December 2026, then $1.50 and $7.50 from 1 January 2027. Gemini 4 Argon, Google's flagship, is $2 and $10 through 30 November 2026, then $4 and $20 from 1 December. Batch halves every rate. Prices read on Google's pricing page, October 2026.

Google is the only one of the three big API vendors that prints an expiry date next to its prices, and that date is the story of this page for anyone budgeting past November. A budget built on today's Flash rate is wrong by a factor of two in nine weeks, and a budget built on Argon is wrong in five. Our view: the Gemini API is the cheapest mid tier on the market until the end of November and merely level with Claude and OpenAI after it.

Every price here was read on Google's Gemini API pricing page and its rate limits page on 26 October 2026, in US dollars per million tokens. The consumer plans (AI Plus, Pro and Ultra) are a separate bill, on Gemini pricing; the 20-model cross-vendor sheet is LLM API prices 2026.

Toolradar data: the AI assistants ranking on Toolradar, the software directory we run, evaluated 1,574 tools as of October 2026; the product records are Google Gemini, Anthropic API and OpenAI API. AIpresso data: this newsletter reaches 29,000+ readers as of November 8, 2026, with an open rate of 31%.

Gemini API prices per million tokens, October 2026

All figures are US dollars per million tokens at the paid tier. "Now" is the price printed today; "Later" is the price Google prints next to it with a start date.

Model	Input now	Output now	Input later	Output later	Date of change	Batch input now	Batch output now	Free tier
Gemini 4 Argon	2.00	10.00	4.00	20.00	1 December 2026	1.00	5.00	Not available
Gemini 3.8 Flash	0.75	3.75	1.50	7.50	1 January 2027	0.375	1.875	Free of charge
Gemini 3.6 Flash	0.75	3.75	1.50	7.50	1 January 2027	0.375	1.875	Free of charge
Gemini 3.5 Flash-Lite	0.30	2.50	No change printed			0.15	1.25	Free of charge
Gemini 3.1 Flash-Lite	0.25 (0.50 audio)	1.50	No change printed			0.125	0.75	Free of charge
Gemini 3.1 Pro Preview, prompts up to 200,000 tokens	2.00	12.00	No change printed			1.00	6.00	Not available
Gemini 3.1 Pro Preview, prompts over 200,000 tokens	4.00	18.00	No change printed			2.00	9.00	Not available
Gemini 3 Flash Preview	0.50 (1.00 audio)	3.00	No change printed			0.25	1.50	Free of charge
Gemini 2.5 Pro, prompts up to 200,000 tokens	1.25	10.00	No change printed			0.625	5.00	Free of charge
Gemini 2.5 Flash	0.30 (1.00 audio)	2.50	No change printed			0.15	1.25	Free of charge
Gemini 2.5 Flash-Lite	0.10 (0.30 audio)	0.40	No change printed			0.05	0.20	Free of charge

Output prices include thinking tokens on every model. The Flex tier is priced the same as Batch on every row, and the Priority tier costs 1.8 times standard (Argon at $3.60 and $18 today, 3.8 Flash at $1.35 and $6.75). Gemini 4 Argon and 3.1 Pro Preview have no free tier; the Flash and Flash-Lite models are free of charge on the free tier with rate limits, and Google's page marks free-tier data as used to improve its products.

What the two increases add to a 100-million-token month

Take a product that sends 80 million input tokens and receives 20 million output tokens a month at standard rates, 100 million tokens in total. Our arithmetic on the printed prices:

Model	Monthly cost now	Monthly cost after the change	Increase	Batch cost now	Batch cost after
Gemini 4 Argon	360	720 from 1 December 2026	360	180	360
Gemini 3.8 Flash	135	270 from 1 January 2027	135	67.50	135
Gemini 3.6 Flash	135	270 from 1 January 2027	135	67.50	135
Gemini 3.5 Flash-Lite	74	No change printed	0	37	37
Gemini 3.1 Flash-Lite	50	No change printed	0	25	25
Gemini 3.1 Pro Preview (short prompts)	400	No change printed	0	200	200
Gemini 3 Flash Preview	100	No change printed	0	50	50
Gemini 2.5 Pro (short prompts)	300	No change printed	0	150	150
Gemini 2.5 Flash	74	No change printed	0	37	37
Gemini 2.5 Flash-Lite	16	No change printed	0	8	8

Figures in US dollars a month. The increases are exactly 2x, so a $360 Argon month becomes $720 and a $135 Flash month becomes $270; a team on ten times that traffic is looking at $3,600 more a month on Argon from December. The unchanged rows are the escape routes: 3.5 Flash-Lite at $74 keeps most of 3.8 Flash's price advantage after January without a printed end date, and 2.5 Flash-Lite at $16 is the cheapest text model on Google's page.

The other thing the table shows is that after its increase, 3.8 Flash at $270 lands above 3.1 Pro Preview's input price and close to 2.5 Pro's $300, so "Flash is the cheap one" stops being true on 1 January 2027 for anyone who did not move down a tier.

How is the Gemini API billed?

Per token, with three meters and one tier ladder that decides how fast you are allowed to spend. Input tokens cover the prompt, history and any file or audio you attach, and audio input costs more on the Flash-Lite and older Flash models. Output tokens include the model's thinking, so a reasoning-heavy request on Argon pays the full output rate for tokens you never see. Context caching bills cached prefix reads at a tenth of the input price, $0.10 on Argon and $0.075 on 3.8 Flash today; Argon supports implicit caching only, while the Flash and Pro models also allow explicit caches that carry a storage charge of $0.50 per million tokens per hour on 3.8 Flash (rising to $1 in January) and $4.50 on 3.1 Pro Preview.

The tier ladder is the part newcomers miss. The free tier needs no billing account and sends Flash traffic for $0 with rate limits. Tier 1 starts when you link a billing account and carries a $250 billing cap and a spend-based limit of $10 per rolling 10 minutes. Tier 2 opens after $100 paid plus three days, with a $2,000 cap and $50 per 10 minutes; Tier 3 after $1,000 paid plus 30 days, with caps from $20,000 and $200 per 10 minutes. A launch that spends faster than its tier allows gets 429 RESOURCE_EXHAUSTED errors, not a bigger bill, so the spend limits are a ceiling to plan around rather than a surprise.

AIpresso readers get the December and January rate changes in the daily brief when they land; the signup is at the end of this page.

Grounding, batch and the other modifiers

Modifier	Effect	Applies to
Batch	50% off input and output	Every model with a Batch tab
Flex	Same prices as Batch	Argon, Flash, Flash-Lite, Pro models
Priority	1.8x standard	Same models
Grounding with Google Search	5,000 free requests a month shared across Gemini 3 and newer models, then $14 per 1,000	Gemini 3 and newer
Grounding with Google Search, 2.5 models	1,500 requests a day free, then $35 per 1,000 grounded prompts	Gemini 2.5 Pro, Flash, Flash-Lite
Grounding with Google Maps	5,000 free a month, then $14 per 1,000	Gemini 3 and newer
Context caching	0.1x input price on reads; storage per million tokens per hour on explicit caches	Varies by model
Long prompts	Higher rate above 200,000 tokens	3.1 Pro Preview, 2.5 Pro

One request can fire more than one search query, and Google bills each query, so a grounded assistant with a few thousand users can leave the 5,000 free requests behind in a day. At $14 per 1,000 that is still cheap next to the tokens the search results add to the prompt, which are billed at the model's input rate.

Gemini vs Claude vs OpenAI on the same workloads

Prices for the two rivals were read the same day on Anthropic's pricing page and OpenAI's pricing page. First, the 100-million-token month from above, at standard rates:

Tier	Gemini	Claude	OpenAI
Small	3.8 Flash 135 (270 from January); 3.5 Flash-Lite 74	Haiku 5.5 18	gpt-6-luna 18
Mid	4 Argon 360 (720 from December)	Sonnet 5.5 360	gpt-6.1-sol 360
Flagship	4 Argon is Google's top model	Opus 5.5 720	gpt-6-astra 1,800

Figures in US dollars a month. The small tier is where Google is dearest: Haiku 5.5 and gpt-6-luna both price at $0.10 and $0.50, so the same month costs $18 against $135 on 3.8 Flash, and even 2.5 Flash-Lite's $16 only matches them with an older model. The mid tier is a three-way tie at $360 until 30 November, after which Argon at $720 costs what Claude Opus 5.5 does and twice what Sonnet 5.5 or gpt-6.1-sol do.

Second, a cached chatbot: 100,000 conversations a month with a 1,500-token cached system prompt and 500 fresh tokens per message, so 50 million fresh input tokens, 150 million cache-read tokens and 30 million output tokens.

Model	Fresh input	Cache reads	Output	Monthly total
Gemini 4 Argon (through 30 November)	100	15	300	415
Gemini 3.8 Flash (through 31 December)	37.50	11.25	112.50	161.25
Gemini 3.5 Flash-Lite	15	4.50	75	94.50
Claude Sonnet 5.5	100	15	300	415
Claude Haiku 5.5	5	1.50	15	21.50
gpt-6.1-sol	100	15	300	415
gpt-6-luna	5	1.50	15	21.50

Our arithmetic, in US dollars a month, with cache reads at each vendor's printed cached-input rate and the one-off storage charges ignored. The ranking does not move: Google's mid model ties the rivals' mid models to the dollar, and Google's small models cost four to eight times the rivals' small models. Anthropic's rates by model are on Claude API pricing; which assistant does better work is the Claude vs Gemini and Gemini vs ChatGPT comparisons, and the Gemini review covers the product.

Is the Gemini API free?

Yes, for the Flash and Flash-Lite models and 2.5 Pro: the free tier prints "Free of charge" for input and output on those rows, with rate limits per project and no billing account required, and Google AI Studio is the place to use it. Gemini 4 Argon and 3.1 Pro Preview show "Not available" on the free tier, so the flagship costs money from the first token. The free tier also allows Google to use your prompts to improve its products, which the paid tier does not, so it is a prototyping tier, not a production one.

Where the Gemini bill surprises people

• The printed end dates. Argon doubles on 1 December 2026 and both current Flash models on 1 January 2027, and nothing on the page says what happens to the "later" price after that.
• Thinking is output. Every output price on the page is marked "including thinking tokens".
• Long prompts cost more on the Pro models. 3.1 Pro Preview goes from $2 and $12 to $4 and $18 past 200,000 tokens, and 2.5 Pro doubles its input rate and adds half again to its output rate.
• Audio input is priced separately on 3.1 Flash-Lite, 3 Flash Preview, 2.5 Flash and 2.5 Flash-Lite, at two to three times the text rate.
• Explicit caches rent storage by the million tokens per hour, so a cache you forget to delete keeps billing.
• Spend caps stop traffic. Tier 1's 10-minute spend limit is about $43,000 a month if you could spend it continuously, but a traffic spike hits the 10-minute window first.
• Grounding counts queries, not prompts, on Gemini 3 and newer; the 2.5 models bill per grounded prompt at $35 per 1,000.

Which Gemini model should you pay for?

3.8 Flash until 31 December, then decide. At $0.75 and $3.75 it is the best price-to-capability row on Google's page today, and it is free to prototype. Calendar the January doubling and test 3.5 Flash-Lite on your traffic before then, because at $0.30 and $2.50 it keeps most of the saving with no end date printed.

Argon for the hardest work, priced month by month. It ties Claude Sonnet 5.5 and gpt-6.1-sol to the dollar until 30 November, which makes it the cheapest flagship from any vendor for five more weeks; from December it costs what Opus 5.5 does. Skip Google's small tier if the small tier is the whole product: Haiku 5.5 and gpt-6-luna do the same 100-million-token month for $18. And for one person rather than a product, the consumer plans on ChatGPT pricing and Claude pricing are a different, flat bill.

Methodology and sources

Every per-model price, the "later" prices and their start dates, batch, flex and priority rates, caching and storage charges, grounding allowances and free-tier availability were read on Google's Gemini API pricing page on 26 October 2026. Usage tiers, billing caps and spend-based limits come from the rate limits page the same day. Claude and OpenAI rates were read the same day on their pricing pages linked above, at standard tiers.

The 100-million-token month and the cached chatbot are our workloads: the token counts are assumptions stated in each section, the dollar figures are those counts multiplied by the printed rates, and token counts are treated as equal across vendors although tokenizers differ. The Toolradar count is from its AI assistants category as of October 2026.

Louis Corneloup, founder of Toolradar and Dupple, wrote this page; we run no test lab and give no rating. The method is on how we rate, corrections go through the editorial policy, and the author note is on about. Google's page carries dates on its prices for a reason; re-read it on 1 December and 1 January.

FAQ

How much does the Gemini API cost per million tokens in 2026?

Gemini 4 Argon costs $2 per million input tokens and $10 per million output tokens through 30 November 2026, then $4 and $20. Gemini 3.8 Flash costs $0.75 and $3.75 through 31 December 2026, then $1.50 and $7.50. Gemini 3.5 Flash-Lite is $0.30 and $2.50 and 2.5 Flash-Lite $0.10 and $0.40, with no change printed. Batch halves every rate.

Is Gemini 4 Argon available on the free tier?

No. Google's page prints "Not available" for Argon and for 3.1 Pro Preview on the free tier. The Flash and Flash-Lite models and 2.5 Pro are free of charge on the free tier with rate limits, and the free tier's data may be used to improve Google's products.

When do Gemini API prices go up?

Gemini 4 Argon rises from $2 and $10 to $4 and $20 per million tokens on 1 December 2026, and Gemini 3.8 Flash and 3.6 Flash rise from $0.75 and $3.75 to $1.50 and $7.50 on 1 January 2027. Both are printed on the pricing page next to the current rates. On an 80-million-input, 20-million-output month that is $360 more on Argon and $135 more on Flash, our arithmetic.

Is the Gemini API cheaper than Claude or OpenAI?

In the mid tier it is level: Argon, Claude Sonnet 5.5 and gpt-6.1-sol all cost $360 on a 100-million-token month until Argon doubles in December. In the small tier it is dearer: 3.8 Flash costs $135 on the same month against $18 on Claude Haiku 5.5 or gpt-6-luna. Google has no model priced like Fable 5.1 or gpt-6-astra.

How does Gemini batch pricing work?

Every Batch tab on the page is half the standard rate for input and output: Argon at $1 and $5 today, 3.8 Flash at $0.375 and $1.875, 3.5 Flash-Lite at $0.15 and $1.25. The Flex tier carries the same prices as Batch, and the scheduled increases apply to the batch rates too.

What does grounding with Google Search cost?

On Gemini 3 and newer models, 5,000 search requests a month are free, shared across those models, then $14 per 1,000 requests; one prompt can trigger several requests and each is billed. On the 2.5 models the allowance is 1,500 requests a day, then $35 per 1,000 grounded prompts. Tokens from the search results are billed as input at the model's rate.

Cite this: AIpresso, "Gemini API Pricing 2026: Flash, Argon and the 2027 Increase", October 2026.

AIpresso is the short daily brief on which AI prices moved and by how much. Subscribe at aipresso.com and get the December and January changes the day they land.

Some offers on this page may be paid placements or contain affiliate links.