Gemini API Pricing Guide 2026: Token Pricing, Free Quotas & Cost Estimation
October 2026 update: Google moved its flagship Gemini 4 Argon to GA on October 13 (no free tier; $2/$10 through 2026-11-30, $4/$20 from 2026-12-01). Gemini 3.7 Flash and 3.5 Flash were deprecated on October 8, 2026; their endpoints now automatically route to 3.8 / 3.6 Flash, and they can no longer be selected in AI Studio. (Sources: Gemini API models, deprecations)
September 2026 update: Google shipped Gemini 3.6 Flash, 3.5 Flash-Lite and 3.5 Flash Cyber on July 21, followed by Gemini 3.8 Flash at the same price on September 2. Meanwhile Gemini 2.0 Flash and 2.0 Flash-Lite were shut down on June 1, 2026. All price tables below have been rebuilt from the official Gemini API pricing page.
"What happens when free quota runs out?" "How much will a month cost?" These are the two most common questions developers ask when getting started with Gemini API. The good news is that Gemini API's free quota is quite generous for small projects; the bad news is that once traffic ramps up, costs might be higher than you imagine.
This article will completely break down Gemini API's pricing model, from token concepts to actual cost estimation, helping you plan your budget. For Gemini's complete product line and pricing structure, refer to Gemini Pricing Complete Guide.
Gemini API Pricing Model Overview
Gemini API uses Token-based pricing—pay for what you use, no monthly fees or subscription fees.
What are Tokens?
Tokens are the basic units AI models use to process text. They're not "characters" or "words," but the smallest segments the model divides text into.
English Token Estimation:
- 4 English letters ≈ 1 token
- 1000-word English article ≈ 1,250 - 1,700 tokens (official: 100 tokens ≈ 60-80 English words)
How Are Tokens Calculated?
Gemini API costs are divided into two parts:
- Input Tokens: Content you send to the API (prompt + context)
- Output Tokens: Content AI replies to you
Output tokens are usually about 4-8 times more expensive than input tokens, because generating content requires more computation than understanding it.
Input vs Output Price Difference
| Item | Description | Price Difference |
|---|---|---|
| Input Tokens | Content you give AI | Cheaper |
| Output Tokens | AI's reply to you | More expensive (about 4-8x) |
Practical Impact: If your application is "input long text, output summary," costs will be much lower than "input question, output long text."
Need Help with API Cost Estimation?
Token pricing looks simple, but actual usage estimation often goes wrong. Let a professional consultant help you evaluate to avoid billing surprises after launch.
Book Architecture Consultation
Gemini API Free Quotas
Google provides quite generous free quotas, friendly for development testing and small projects.
How to Check Your Free Tier (as of September 2026)
One thing worth flagging: Google no longer publishes fixed free-quota numbers. The official rate limits page now states that limits "depend on a variety of factors (such as your usage tier)" and directs developers to check their own account in Google AI Studio.
In other words, any table you find online claiming "Gemini free tier: 15 RPM / 1M tokens per day" is no longer a reliable reference.
The correct approach: check your own rate limits in AI Studio. Quotas vary by account tier (whether billing is attached, historical usage).
Most Gemini text models still offer free-tier tokens, including 3.8/3.6 Flash, 3.5 Flash-Lite, 3.1 Flash-Lite, the 2.5 family and Gemini Embedding (3.7 Flash and 3.5 Flash were deprecated on 2026-10-08). Image/video generation models (3.1 Flash Image, 3 Pro Image), the Pro Preview line, and the flagship Gemini 4 Argon (GA on 2026-10-13) have no free tier.
What Are Free Quotas Suitable For?
| Use Case | Suitability | Description |
|---|---|---|
| Development Testing | Very Suitable | More than enough for testing features |
| Side Project | Suitable | Sufficient for low-traffic applications |
| MVP Validation | Suitable | Validate first, consider paying later |
| Production Environment | Depends on traffic | Low traffic might be enough |
| High-Traffic Applications | Not Suitable | Need paid plan |
Key Point: Free quota limitations are mainly RPM (requests per minute), not total usage. If your application needs to handle many requests simultaneously, free quota quickly becomes insufficient.
Gemini API Paid Price Table
After exceeding free quotas, billing begins.
Main Text Model Price Table (October 2026)
| Model | Input Price | Output Price | Positioning |
|---|---|---|---|
| Gemini 4 Argon | $2.00/1M tokens (through 2026-11-30; $4.00 from 2026-12-01) | $10.00/1M tokens (through 2026-11-30; $20.00 from 2026-12-01) | Flagship, GA on 2026-10-13, no free tier; cached input $0.10 ($0.20 from 12-01), implicit caching only |
| Gemini 3.8 / 3.6 Flash | $0.75/1M tokens (through 2026-12-31; $1.50 from 2027-01-01) | $3.75/1M tokens (through 2026-12-31; $7.50 from 2027-01-01) | Current workhorse (3.8, Stable since 2026-09-02, is the latest; 3.6 is the previous generation); 3.6 uses 17% fewer output tokens than 3.5 Flash |
| Gemini 3.5 Flash | — (deprecated) | — (deprecated) | Previous workhorse, deprecated on 2026-10-08 (was $1.50 / $9.00); its endpoint now routes to 3.8 / 3.6 Flash. 3.7 Flash was deprecated the same day |
| Gemini 3.5 Flash-Lite | $0.30/1M tokens | $2.50/1M tokens | Cheapest in the 3.5 class, high throughput |
| Gemini 3.1 Flash-Lite | $0.25/1M tokens | $1.50/1M tokens | Cheaper still; audio input $0.50; shutdown announced for 2027-05-07 at the earliest, successor is 3.5 Flash-Lite |
| Gemini 3.1 Pro Preview | $2.00 (≤200k) / $4.00 (>200k) | $12.00 (≤200k) / $18.00 (>200k) | High-end reasoning, no free tier |
| Gemini 2.5 Pro | $1.25 (≤200k) / $2.50 (>200k) | $10.00 (≤200k) / $15.00 (>200k) | Previous flagship, still available (existing users only) |
| Gemini 2.5 Flash-Lite | $0.10/1M tokens | $0.40/1M tokens | Lowest unit price across the line (existing users only) |
Prices in USD. Source: official Gemini API pricing page (Gemini 4 Argon and the 3.7 / 3.5 Flash deprecations, with endpoints auto-routing to 3.8 / 3.6 Flash, added 2026-10-28). Google may adjust at any time.
⚠️ Retired models: Gemini 3.7 Flash and 3.5 Flash were deprecated on October 8, 2026 (endpoints route to 3.8 / 3.6 Flash); Gemini 2.0 Flash and 2.0 Flash-Lite were shut down on June 1, 2026; Veo 2 and Veo 3 on June 30; Imagen 4 on August 17. If your code still hardcodes 2.0 Flash, Veo 2/3, or Imagen 4, those calls already fail; code that hardcodes 3.7 / 3.5 Flash still works for now (requests are auto-routed), but responses come from the newer models, so update the model string directly.
Model Characteristics
Gemini 4 Argon
- The current flagship, GA on 2026-10-13, with no free tier — testing bills from the first token
- Limited-time pricing: $2/$10 per 1M through 2026-11-30, then $4/$20 from 2026-12-01; budget for the latter
- Implicit caching only; cached input is $0.10, rising to $0.20 from 12-01
Gemini 3.6 Flash
- Google called it the workhorse at its July 2026 launch; 3.8 Flash at the same price followed in September (3.7 Flash, also the same price, was deprecated on October 8)
- Uses roughly 17% fewer output tokens than 3.5 Flash, so the same task costs less
- Suitable for: most production workloads, agent applications
Gemini 3.5 Flash-Lite
- Fastest and cheapest of the 3.5 class; Google cites 350 output tokens/second per the Artificial Analysis Index
- Suitable for: high-throughput, cost-sensitive batch processing
Gemini 2.5 Flash-Lite
- Lowest unit price on the entire line ($0.10 / $0.40)
- Suitable for: simple classification and tagging
- Existing users only: Google now limits the 2.5 family to accounts that have used it before; for new projects, use 3.5 Flash-Lite or 3.8 Flash
Gemini 3.1 Pro Preview
- High-end reasoning with a separate price tier above 200k context
- No free tier — testing bills from the first token
Gemini 3.5 Flash Cyber
- A security-specialised model fine-tuned to find and fix vulnerabilities
- Not purchasable by general developers: currently limited to governments and trusted partners through CodeMender as a limited-access pilot
Gemini vs OpenAI API Pricing Comparison
This is what developers care most about—who's cheaper, Gemini API or OpenAI API?
Price Comparison Table
| Tier | Gemini | Price (in/out) | OpenAI | Price (in/out) |
|---|---|---|---|---|
| Entry | Gemini 2.5 Flash-Lite (existing users only) | $0.10 / $0.40 | GPT-6 Luna | $0.10 / $0.50 |
| Light | Gemini 3.5 Flash-Lite | $0.30 / $2.50 | GPT-6 Luna | $0.10 / $0.50 |
| Workhorse | Gemini 3.8 Flash | $0.75 / $3.75 (promotional through 2026-12-31; $1.50 / $7.50 from 2027-01-01) | GPT-6 Luna | $0.10 / $0.50 |
| High-end | Gemini 3.1 Pro Preview | $2.00 / $12.00 | GPT-6.1 Sol | $2.00 / $10.00 |
| Flagship | Gemini 4 Argon (GA 2026-10-13) | $2.00 / $10.00 (through 2026-11-30; $4.00 / $20.00 from 2026-12-01) | GPT-6 Astra | $10.00 / $50.00 |
OpenAI prices from the official OpenAI API pricing page; GPT-6 Astra launched 2026-09-03, GPT-6 Luna 2026-09-22, GPT-6.1 Sol 2026-09-29.
Price Difference Analysis
| Comparison | Cheaper | Gap |
|---|---|---|
| Entry (2.5 Flash-Lite vs 6 Luna) | Gemini (output) | Same input price; Gemini 20% less on output |
| Light (3.5 Flash-Lite vs 6 Luna) | OpenAI | OpenAI costs about a third on input and a fifth on output |
| Workhorse (3.8 Flash vs 6 Luna) | OpenAI | Gemini costs 650% more on both input and output (1,400% more once unit prices double from 2027-01-01) |
| High-end (3.1 Pro Preview vs 6.1 Sol) | OpenAI (output) | Same input price; Gemini costs 20% more on output |
| Flagship (4 Argon vs 6 Astra) | Gemini | 80% less on both input and output during the limited-time price (through 2026-11-30); still 60% less at $4 / $20 from 2026-12-01 |
Conclusion: The old rule of thumb that "Gemini is roughly 50% cheaper than OpenAI across the board" no longer holds in 2026.
It is now tier-dependent:
- From the entry to the workhorse tier, OpenAI's GPT-6 Luna ($0.10 / $0.50) undercuts both Gemini 3.5 Flash-Lite and 3.8 Flash; against 3.8 Flash ($0.75 / $3.75, promotional through 2026-12-31; $1.50 / $7.50 from 2027-01-01) it costs about 1/7.5 on both input and output. Only Gemini 2.5 Flash-Lite's output price ($0.40) is lower than Luna's, by 20%
- At the flagship tier, Gemini 4 Argon (GA on 2026-10-13; $2 / $10 through 2026-11-30, $4 / $20 from 2026-12-01) is cheaper than GPT-6 Astra ($10 / $50) at both price levels, and new projects can use it; the previous-generation flagship 2.5 Pro is open only to accounts that have used it before
- The high-end tier is close to level: Gemini 3.1 Pro Preview ($2.00 / $12.00) and GPT-6.1 Sol ($2.00 / $10.00) match on input, with OpenAI slightly cheaper on output
So model selection can no longer be settled with "Google is cheaper." It depends which tier your workload actually sits in and whether you can still use the 2.5 family. Whatever the tier, it is worth re-running the numbers.
Performance vs Cost Trade-off
Cheaper isn't necessarily better. When choosing, consider:
| Aspect | Gemini | OpenAI |
|---|---|---|
| Price | Cheaper at the flagship tier (Gemini 4 Argon) and with the 2.5 family (existing users only) | Mostly cheaper from entry to workhorse tier |
| Ecosystem | Newer | More mature |
| Documentation | Medium | Rich |
| Third-party Integration | Fewer | Very many |
| Chinese Quality | Medium | Better |
If your project is cost-sensitive, re-run the numbers for your actual tier on both sides; if you need rich third-party tool integration, OpenAI's ecosystem is more complete.
Not Sure Which API to Choose?
Gemini, OpenAI, Claude, Azure... so many API choices, each with pros and cons. Let experts recommend the best combination based on your application scenario.
Cost Estimation Examples
Finished with theory, let's look at real cases.
Example 1: Chatbot (1000 conversations/day)
Assumptions:
- Each conversation: 500 input tokens, 300 output tokens
- 1000 conversations daily
- Using Gemini 3.5 Flash-Lite
Calculation:
- Daily input: 500 × 1000 = 500,000 tokens
- Daily output: 300 × 1000 = 300,000 tokens
- Daily cost: (0.5M × $0.30) + (0.3M × $2.50) = $0.15 + $0.75 = $0.90
- Monthly cost: $0.90 × 30 = $27 (about NT$864)
Example 2: Document Summary Service (100 documents/day)
Assumptions:
- Each document: 10000 input tokens, 500 output tokens
- 100 documents daily
- Using Gemini 3.1 Pro Preview (quality requirements)
Calculation:
- Daily input: 10000 × 100 = 1 million tokens
- Daily output: 500 × 100 = 50,000 tokens
- Daily cost: (1M × $2.00) + (0.05M × $12.00) = $2.00 + $0.60 = $2.60
- Monthly cost: $2.60 × 30 = $78 (about NT$2,496)
Example 3: Code Generation Tool (500 requests/day)
Assumptions:
- Each request: 800 input tokens, 1500 output tokens
- 500 requests daily
- Using Gemini 3.6 Flash (code quality)
Calculation:
- Daily input: 800 × 500 = 400,000 tokens
- Daily output: 1500 × 500 = 750,000 tokens
- Daily cost: (0.4M × $0.75) + (0.75M × $3.75) = $0.30 + $2.81 = $3.11
- Monthly cost: $3.11 × 30 = $93.3 (about NT$2,986; unit prices double from 2027, to about $186.75)
NT$ figures assume roughly NT$32 per USD; actual cost varies with the exchange rate and card fees.
Cost Calculation Formula
Monthly Cost = (Daily Input Tokens × Input Unit Price + Daily Output Tokens × Output Unit Price) × 30
Compared with Gemini prices before 2025, these costs are noticeably higher — Gemini's unit prices moved up across the line after the 1.5 generation. If your budget still uses 2025 figures, re-run it.
Tips for Reducing API Costs
Cost estimation done, now how to save money.
1. Prompt Optimization to Reduce Tokens
Bad Prompt (wastes tokens):
Please act as a very professional article summarization expert,
you need to carefully read the following article,
then use your professional knowledge,
to organize the key points of the article...
Good Prompt (concise):
Summarize the following article, list 3 key points:
2. Choose the Right Model
| Task Type | Recommended Model | Reason |
|---|---|---|
| Simple classification, tagging | Gemini 2.5 Flash-Lite (existing users only; new projects use 3.5 Flash-Lite) | Lowest unit price on the line |
| High-throughput batch work | Gemini 3.5 Flash-Lite | Fast and cheap |
| General conversation, agents | Gemini 3.8 Flash (or 3.6 Flash at the same price) | Current workhorse, same price as 3.6 |
| Complex reasoning | Gemini 3.1 Pro Preview | Quality requirements |
| Flagship-level reasoning | Gemini 4 Argon | Flagship, GA on 2026-10-13; no free tier, $2 / $10 through 2026-11-30 and $4 / $20 from 12-01 |
| Long context (>200k) | Gemini 2.5 Pro (existing users only) | Cheaper long-context tier |
3. Caching Strategy
If the same questions will repeat, consider:
- Cache common question answers
- Use vector database to store similar questions
- Set TTL for periodic updates
4. Batch Processing
Combine multiple small requests into one large request:
- Reduce API call count
- Lower network latency
- But watch context length limits
Vertex AI (Now Gemini Enterprise Agent Platform) vs AI Studio
There are two ways to use Gemini API, with slightly different pricing and features.
Two Access Methods
| Item | AI Studio | Vertex AI |
|---|---|---|
| Positioning | Developer / Testing | Enterprise Production Environment |
| Setup Complexity | Simple | More Complex |
| Billing Method | API Key Direct Billing | GCP Billing Integration |
| Free Quota | More | Less |
| SLA | None | Yes |
| Security | Basic | Enterprise-grade |
Price Differences
Agent Platform (formerly Vertex AI) pricing may differ from AI Studio (the official pricing page notes the two can differ), but provides:
- Enterprise-grade SLA
- Better security and compliance
- GCP integration (VPC, IAM)
- Volume discounts
Selection Recommendations
| Scenario | Recommendation |
|---|---|
| Personal Projects | AI Studio |
| Small Startups | AI Studio |
| Enterprise Production | Vertex AI |
| Need SLA | Vertex AI |
| Already Have GCP | Vertex AI |
If you're a developer who also wants to learn about Google's code assistant tools, refer to Gemini Code Assist Pricing and Feature Review.
Frequently Asked Questions FAQ
What Happens When Free Quota is Exceeded?
On the free tier, requests are blocked once you hit the rate limits; nothing is billed automatically. Billing starts only after you link a billing account and prepay (Prepay, minimum $5) to move to a paid tier, and the API stops serving requests when the prepaid balance runs out. Recommendations:
- Set usage alerts
- Set budget limits
- Link payment method to avoid service interruption
How to Monitor API Usage?
In Google Cloud Console you can view:
- Real-time usage charts
- Usage by model
- Cost estimates
You can also query remaining quota via API.
Are There Enterprise Contract Discounts?
Yes. Google's Enterprise plan offers usage-based volume discounts; contact Google sales to negotiate.
How to Read API Bills?
In Google Cloud Console → Billing → Reports you can see:
- Costs by service
- Cost trends over time
- Cost forecasts
It's recommended to set daily/monthly budget alerts to avoid unexpected overspending.
Conclusion: API Cost Planning Recommendations
Development Stage
- Use free quota first: More than enough for testing
- Choose the right model: Test with Flash first, switch to Pro when needed
- Optimize prompts: Reduce unnecessary tokens
Launch Stage
- Set budget alerts: Avoid bill explosions
- Monitor actual usage: Compare with estimates
- Consider caching: Reduce repeated calls
Scaling Stage
- Negotiate enterprise discounts: High volume can negotiate prices
- Evaluate Agent Platform (formerly Vertex AI): Upgrade if you need SLA
- Mixed models: Different tasks use different models
Need API Architecture Consultation?
API cost planning isn't just looking at price tables—you also need to consider architecture design, caching strategy, model selection. Let professional consultants help you design the optimal solution.
Book Cost Optimization Consultation
Further Reading
- Specs, the migration checklist, and cost math for the new flagship, see Gemini 4 Argon Guide
- Return to complete pricing guide, see Gemini Pricing Complete Guide
- Developer tool review, see Gemini Code Assist Pricing and Features
- More detailed comparison with ChatGPT API, see Gemini vs ChatGPT Pricing Comparison
- Consumer version analysis, see Google AI Pro (formerly Gemini Advanced) Complete Feature Review
References
- Gemini API Official Pricing
- Gemini API Rate Limits (free quota lookup)
- Gemini API models (Gemini 4 Argon GA)
- Gemini API deprecations (3.7 / 3.5 Flash deprecated 2026-10-08)
- Google announcement: Gemini 3.6 Flash, 3.5 Flash-Lite, 3.5 Flash Cyber
- Gemini Enterprise Agent Platform (formerly Vertex AI) Pricing
- OpenAI API Pricing
- Google Cloud Billing Documentation
Need Professional Cloud Advice?
Whether you're evaluating cloud platforms, optimizing existing architecture, or looking for cost-saving solutions, we can help
Book Free ConsultationRelated Articles
Gemini Pricing Complete Guide 2026: Free vs Paid Version Differences, API Pricing Full Analysis
Complete analysis of Google Gemini pricing: free tier limits, Google AI Plus/Pro/Ultra Taiwan monthly fees (Gemini Advanced has been renamed), and API token pricing.
AI APIWhat Is Gemini API? 2026 Complete Guide to Google Gemini API Integration, Pricing & Development
The most comprehensive 2026 Gemini API development guide. Detailed coverage of Google Gemini API application process, Python integration tutorial, model version comparison, token pricing, and enterprise use cases to help you get started with AI development fast.
GeminiIs Gemini Advanced Worth It? (Now Renamed Google AI Pro) 2026 Complete Feature Review and Cost Analysis
The name "Gemini Advanced" has been retired — it is now Google AI Pro at NT$650/month with Gemini 3.1 Pro, 4x usage, and 5 TB of storage. A full review of features, cost, and who it's for, with real usage notes and buying advice.