Gemini API Pricing (October 2026): Argon, Pro, Flash & Flash-Lite Costs

The Gemini API charges per million tokens (MTok), with a free tier for many Flash and Flash-Lite models. As of October 17, 2026, Google's newest model, Gemini 4 Argon, costs $2 input / $10 output on introductory pricing that doubles on December 1, 2026. Gemini 3.1 Pro Preview costs $2 / $12, Gemini 3.8 Flash costs $0.75 / $3.75 until it doubles on January 1, 2027, and Gemini 3.5 Flash-Lite costs $0.30 / $2.50. Several Gemini prices are scheduled to change within the next three months, and the worked examples below show what that means for your bill.

Current model prices (paid tier, Standard)

Output prices include thinking tokens. Gemini models think before answering by default, so a short visible answer can still use many output tokens. On Gemini 4 Argon, you can choose a thinking level of low, medium (the default) or high. Lower levels use fewer tokens and cost less.

Gemini 4 Argon

Argon is Google's latest and most capable model, aimed at complex coding, agentic work, math and science. It has a single 1M-token window shared by input, thinking and output. Its introductory price of $2 / $10 runs through November 30, 2026, and becomes $4 / $20 from December 1. Argon has no separate long-context price, so a 500K-token prompt costs the same per token as a 5K one. Caching is implicit only (no explicit cache creation), with cached input at $0.10. Argon is not on the free tier.

Gemini 3.1 Pro Preview

Gemini 3.1 Pro is still listed as a preview. It costs $2 / $12 up to 200K input tokens and $4 / $18 above that, with explicit context caching at $0.20 per 1M tokens plus $4.50 per 1M tokens per hour of storage. Its input price is the same as Argon's introductory price and its output is higher, so compare the two on your own prompts. After December 1, Argon will cost more per token than 3.1 Pro for prompts under 200K.

Flash and Flash-Lite

Flash models are the workhorses for most apps: fast, cheap and on the free tier. Gemini 3.8 Flash and 3.6 Flash share the same price and the same increase on January 1, 2027. Flash-Lite models (3.5 and 3.1) are cheaper still and suit high-volume classification, extraction and routing. Gemini 2.5 Flash-Lite, at $0.10 / $0.40, is the cheapest text model on the list, but it is an older generation.

Processing tiers: Standard, Batch, Flex, Priority and Contributor

Batch is for asynchronous jobs, and Flex gives the same half-price rates on regular requests in exchange for lower priority. Priority costs 1.8x Standard for more reliable latency. The Contributor tier is a private, invite-only preview: according to Google's data logging page, it gives up to 50% off on select models (Gemini 4 Argon, 3.1 Pro Preview, 3.8 Flash and 3.5 Flash-Lite) in exchange for letting Google use that data to improve its products. Don't send sensitive or personal data on that tier.

The free tier

Many models are "Free of charge" for input and output on the free tier, including Gemini 3.8 Flash, 3.6 Flash, 3.5 and 3.1 Flash-Lite, the 2.5 models, Gemini Embedding 2 and Gemma 4. Gemma 4 is only available on the free tier, with no paid option. Gemini 4 Argon, Gemini 3.1 Pro Preview and the image-generation models are not free. Google AI Studio itself is free to use in available regions.

• Rate limits are set per project and vary by model. Google doesn't publish a table of free-tier limits; you see yours in Google AI Studio.
• Data use: on the free tier, your prompts and responses may be used to improve Google's products. Paid tiers (except Contributor) are not used this way.
• Grounding with Google Search is not available on the free tier for Gemini 3 and newer models.
• For setup steps, see our Gemini API key guide and the Gemini API free tier page.

Billing, prepay and usage tiers

To move to the paid tier, open Google AI Studio, choose Set up billing on the API keys or Projects page, link a Cloud Billing account and prepay at least $5. The billing docs explain the two payment modes:

• Prepay: buy $5 to $5,000 of credit in advance. Usage comes out of the balance, and at $0 requests fail with HTTP 402. Unused credit expires after 12 months and is non-refundable (except when switching to Postpay).
• Postpay: costs accrue and are charged monthly or when you reach your tier's spend cap. You can switch from Postpay to Prepay but not the other way.

Tier caps are hard stops. When a billing account reaches its tier cap, service pauses for all linked projects until the 1st of the next month. A new Tier 1 account with a production app that costs more than $250 a month will be cut off partway through the month.

Projects can also set an experimental spend cap in AI Studio's Spend page. Google notes that charges can overshoot by up to about 10 minutes of usage before the cap takes effect.

Worked cost examples

These use paid-tier Standard prices and ignore thinking-token variation, which you should measure on your own prompts.

1. A chatbot handling 100,000 requests a month

Each request has 2,000 input tokens and 500 output tokens: 200M input and 50M output tokens a month.

If you build on 3.8 Flash now, budget for the bill doubling in January. Every option except Flash-Lite is also above the $250 Tier 1 cap, so a new billing account will pause before the month ends until it reaches Tier 2.

2. A 500K-token document on Pro vs Argon

You send a 500,000-token document and get an 8,000-token summary. On Gemini 3.1 Pro the prompt is over 200K, so: 0.5 × $4 = $2.00 input plus 0.008 × $18 = $0.144 output, for $2.14. On Gemini 4 Argon at the introductory price: 0.5 × $2 = $1.00 plus 0.008 × $10 = $0.08, for $1.08. From December 1, Argon would be 0.5 × $4 + 0.008 × $20 = $2.16, about the same as 3.1 Pro.

3. Context caching on Gemini 3.1 Pro

You ask 200 questions about the same 100,000-token manual within an hour. Without caching, input costs 200 × 0.1M × $2 = $40. With explicit caching, the 200 cached reads cost 200 × 0.1M × $0.20 = $4, and an hour of storage costs 0.1M × $4.50 = $0.45, so about $4.45 plus the cost of creating the cache. Storage is charged per hour, so delete caches you no longer need.

4. Classifying 1 million documents with Batch

Each document is 800 input tokens with a 50-token label, on Gemini 3.5 Flash-Lite: 800M input and 50M output. Standard: 800 × $0.30 = $240 plus 50 × $2.50 = $125, for $365. Batch: 800 × $0.15 = $120 plus 50 × $1.25 = $62.50, for $182.50. Running the same job on the free tier may be possible in small daily chunks, but free-tier rate limits make a million documents impractical.

5. Search grounding

An app on Gemini 3.8 Flash makes 20,000 grounded requests a month. The first 5,000 are free, and the other 15,000 cost 15 × $14 = $210, on top of the token costs. The 5,000 free requests are shared across all Gemini 3 and newer models on the account.

Tools, media and embeddings

PDF pages are billed at the image token rate. Gemini 2.5 Flash Image (the original Nano Banana) was deprecated and shut down on October 2, 2026.

Credits that can cover Gemini API costs

• Google AI Pro and Ultra: subscribers with Cloud Billing enabled can get monthly Google Cloud credits through the Google Developer Program, which can pay for Gemini API usage. Google's plan comparison lists $10/month for AI Pro and $40 or $100/month for the Ultra tiers. On Prepay, you need a paid balance above $0 to activate them. See Google AI plans pricing.
• Google Cloud $300 trial: for Google Cloud accounts opened after March 2, 2026, the Welcome credit can't be used on Gemini API or AI Studio. It still covers other Google Cloud services (Google Cloud free trial).
• Startups: the Google for Startups Cloud Program gives Cloud credits to eligible startups.

How Gemini compares

Gemini's biggest advantage is its free tier on current Flash models, which OpenAI doesn't offer. At the top end, Gemini 4 Argon's introductory $2 / $10 matches OpenAI's GPT-6.1 Sol price (OpenAI API pricing), but from December 1 it will cost twice that. Compare with Claude API pricing too, and test each model on your real prompts, because token counts and thinking length differ between providers. For other free options, see our list of free AI API keys.

Upcoming and recent price changes

• March 2, 2026: Gemini API usage excluded from the Google Cloud Free Trial credit for new accounts.
• October 2, 2026: Gemini 2.5 Flash Image shut down.
• December 1, 2026: Gemini 4 Argon doubles to $4 / $20 (Batch to $2 / $10, cached input to $0.20).
• January 1, 2027: Gemini 3.8 Flash and 3.6 Flash double to $1.50 / $7.50, along with their caching prices and the 3.8 Flash TTS models.

Google's pricing page was last updated on October 16, 2026. Because several promotional prices end soon, check it again before committing to a model for a long-running project.