Gemini API Pricing 2026: Every Model, Free Tier and Costs

Last Updated: November 1, 2026

Gemini API pricing runs from $0.25 per million input tokens for Gemini 3.1 Flash-Lite to $4 per million for Gemini 4 Argon, Google's flagship, once its introductory rate ends on December 1, 2026. Most Flash and Flash-Lite models also have a free tier, while Argon and Gemini 3.1 Pro are paid only. Every price below comes from Google's official pricing pages, checked on November 1, 2026.

This guide covers per-token prices for every current text model, the free tier and its data trade-off, rate limit tiers, Batch and caching discounts, Google Search grounding, image, video and voice models, how Vertex AI pricing differs, and how Gemini stacks up against OpenAI, Claude, Grok and DeepSeek. It ends with worked cost examples and a plain recommendation on which model to use.

On the paid tier, Gemini 4 Argon costs $2 per million input tokens and $10 per million output tokens until November 30, 2026, then $4 and $20. Gemini 3.8 Flash costs $0.75 and $3.75 until December 31, 2026, then $1.50 and $7.50. Flash-Lite models cost $0.25 to $0.30 for input.

Google bills input and output separately, and output prices include the model's "thinking" tokens. The table shows Standard paid-tier rates from Google's Gemini API pricing page for text, image and video input.

Three details trip people up. First, Argon and the two newest Flash models are on introductory pricing, so today's budgets double in December or January. Second, among the text models above, only 3.1 Pro and 2.5 Pro carry a 200k-token price break. Argon has a 1M-token context window, per Google's models page, and no listed long-prompt surcharge. Third, audio input often costs more, such as $0.50 versus $0.25 per million on Gemini 3.1 Flash-Lite. The 2.5 models remain available, but Google recommends 3.5 Flash-Lite or 3.8 Flash for new projects.

Yes, partly. Gemini 3.8 Flash, 3.6 Flash, 3.5 Flash-Lite, 3.1 Flash-Lite, Gemini 2.5 Pro, Gemini 2.5 Flash and the text-to-speech models can be used free of charge within rate limits. Gemini 4 Argon, Gemini 3.1 Pro Preview, Nano Banana Pro and Gemini Omni Flash are not available on the free tier.

New accounts start on the free tier with no billing set up, per Google's billing docs. Context caching is not offered on it, and Grounding with Google Search on Gemini 3 and newer models is "not available" on the free tier, though you can test it in Google AI Studio. Gemini 2.5 Flash is the exception, with up to 500 grounded requests per day free.

Google does not publish fixed free-tier limits. Its rate limits page says they vary by model and points users to their AI Studio dashboard. Daily quotas reset at midnight Pacific time.

You pay for the free tier with data. Under the Gemini API terms, Google uses free-tier prompts and responses to improve its products, and "human reviewers may read, annotate, and process your API input and output." Google tells free users not to submit sensitive, confidential or personal information.

On the paid tier, Google says it does not use prompts or responses to improve its products and logs them for a limited period only to detect policy violations. Developers serving users in the European Economic Area, Switzerland or the UK may only use paid services.

Paid usage is split into three tiers. Linking a billing account and prepaying at least $5 puts you in Tier 1, $100 in cumulative payments plus three days moves you to Tier 2, and $1,000 plus 30 days moves you to Tier 3.

Since March 23, 2026, new users default to prepaid billing: you buy $5 to $5,000 in credits, which expire after 12 months, and requests fail at a zero balance. Hit your tier cap and service pauses until the first of the next month unless you request an increase.

One trap for startups: Google Cloud's $300 welcome credit cannot be used for the Gemini API or AI Studio on billing accounts opened after March 2, 2026, according to the same billing docs.

Batch mode costs half the Standard price on every model that supports it, in exchange for a target turnaround of 24 hours. Context caching charges a much lower rate for reused input, a tenth of the normal input price or less, plus an hourly storage fee on some models.

Google's Batch API docs say jobs run at "50% of the standard cost" and expire after 48 hours. The pricing page lists Flex at the Batch price and Priority at 1.8 times Standard.

Storage fees add up, so caching pays off when the same long prompt, such as a manual or codebase, is reused many times an hour.

For Gemini 3 and newer models, paid accounts get 5,000 grounded search queries a month free, shared across all those models, then pay $14 per 1,000 queries. Grounding with Google Maps has the same allowance and price.

The fine print matters. Google says "you will be charged for each individual search query performed," so one prompt that triggers three searches counts as three. Older Gemini 2.5 models use a different scheme: 1,500 free grounded requests per day, then $35 per 1,000 grounded prompts.

Images run on the Nano Banana family at about $0.034 to $0.24 each. Video now runs on Gemini Omni Flash at roughly $0.10 per second of 720p. Voice is billed per minute or per million audio tokens.

Two older names are gone. Google's models page lists Imagen 4 as shut down, and its video docs say the Veo 3.1 preview models were shut down in the Gemini API on October 22, 2026, with developers told to move to Omni Flash. For embeddings, Google's embeddings docs list the multimodal gemini-embedding-2 and text-only gemini-embedding-001, with batch embedding at 50% of the default price.

For most models, no. Google Cloud's pricing page, which now brands Vertex AI as Agent Platform, lists the same global-endpoint rates as the Gemini API: $2 and $10 for Argon during the promotion, $0.75 and $3.75 for Gemini 3.8 Flash. Regional (non-global) endpoints cost about 10% more.

The Agent Platform pricing page shows Argon at $2.20 input and $11 output on non-global endpoints during the promotion, and Gemini 3.8 Flash at $0.825 and $4.125.

Vertex also sells Provisioned Throughput, which reserves capacity, priced per GSU per hour from $7.14 on a one-week commitment down to $2.74 on a one-year commitment at global rates.

The simple rule: use the Gemini API for the fastest start, and Vertex if your company already runs on Google Cloud or needs regional endpoints and reserved capacity.

Gemini 4 Argon's introductory $2 and $10 matches OpenAI's GPT-6.1 Sol and Claude Sonnet 5.5, and its December price of $4 and $20 matches Claude Opus 5.5. At the budget end, Gemini Flash-Lite is not the cheapest: OpenAI's GPT-6 Luna and Claude Haiku 5.5 both list at $0.10 input and $0.50 output.

Sources: OpenAI API pricing, Anthropic pricing docs, xAI pricing docs and DeepSeek pricing docs, all checked November 1, 2026. For full breakdowns, see our guides to OpenAI API pricing, Claude API pricing, Grok pricing and DeepSeek pricing.

The table hides some catches. Grok 4.7 has the lowest output price of the $2-input flagships at $6, but its rates double once a prompt reaches 200k tokens. Claude Haiku 5.5 jumps to $0.50 and $2.50 for prompts over 100k tokens. DeepSeek is the cheapest capable option outright, and half price again outside its peak hours of 01:00 to 04:00 and 06:00 to 10:00 UTC on weekdays. For the wider trend behind these numbers, read our piece on AI token prices hitting record lows.

A support bot handling 300,000 chats a month costs about $1,013 on Gemini 3.8 Flash today and $2,025 from January. A coding agent running 1,000 large tasks a month on Argon costs $200 in November and $400 from December. All examples use Standard paid-tier list prices.

Assume 300,000 conversations a month, each with 2,000 input tokens and 500 output tokens. That is 600 million input tokens and 150 million output tokens.

Assume 1,000 tasks a month, each with 50,000 input tokens and 10,000 output tokens: 50 million input and 10 million output tokens.

One request with a 300,000-token prompt and 5,000 output tokens is billed at the long-prompt rate: 0.3 x $4 = $1.20, plus 0.005 x $18 = $0.09, for $1.29. Trim the prompt to 180,000 tokens and it costs 0.18 x $2 = $0.36, plus 0.005 x $12 = $0.06, or $0.42.

An app making 20,000 Google Search grounding queries a month on Gemini 3 or newer pays nothing for the first 5,000 and $14 per 1,000 after: 15 x $14 = $210, on top of token costs.

Start with Gemini 3.8 Flash for most production apps, use Flash-Lite for high-volume simple work, and pay for Argon only on hard reasoning and coding tasks.

If you are still choosing a provider, our AI pricing guide lays out what the major AI tools cost side by side.

Paid text model prices run from $0.25 per million input tokens (Gemini 3.1 Flash-Lite) to $4 (Gemini 4 Argon from December 1, 2026), and $1.50 to $20 per million output tokens. Batch mode halves these rates.

Gemini 3.8 Flash and the Flash-Lite models have a free tier with rate limits, but Gemini 4 Argon and Gemini 3.1 Pro Preview are paid only. Free-tier prompts may be used to improve Google's products and read by human reviewers.

Gemini 4 Argon costs $2 per million input tokens and $10 per million output tokens through November 30, 2026, then $4 and $20 from December 1. Batch requests cost half.

Some models do. Gemini 3.1 Pro Preview charges $4 input and $18 output for prompts over 200k tokens, up from $2 and $12, and Gemini 2.5 Pro has a similar break. Google's pricing page lists no long-prompt surcharge for Gemini 4 Argon or the Flash models.

Not on global endpoints, where Vertex AI lists the same per-token rates as the Gemini API. Regional endpoints cost about 10% more, and Vertex adds options such as Provisioned Throughput for reserved capacity.

It depends on the model. Argon's introductory price matches OpenAI's GPT-6.1 Sol at $2 and $10, but GPT-6 Luna at $0.10 and $0.50 undercuts Gemini's Flash-Lite models.

The Gemini API is one of the cheapest ways to start building, thanks to a free tier on its Flash and Flash-Lite models, but the paid prices you see today are partly introductory. Argon doubles to $4 and $20 on December 1, and Gemini 3.8 Flash doubles to $1.50 and $7.50 on January 1, 2027. Budget at those later rates, use Batch and caching where latency allows, and keep sensitive data off the free tier.

By AI Business Weekly Editorial

Written by the AI Business Weekly team with AI assistance, then reviewed, fact-checked against cited sources and edited by Sameer Khan.

About the editor: Sameer Khan is the founder of AI Business Weekly. He has nearly a decade of B2B sales experience across SaaS, research and advisory, and AI, working with HR leaders and executives on research and AI adoption. He holds an MBA from the Ted Rogers School of Management.