Gemini API Pricing 2026: Flash-Lite Costs & TWD Estimates

Gemini API Pricing 2026: Official Rates and TWD Estimates

Gemini API billing depends on the model, the number of input/output tokens, and the service tier. As of 2026-09-24, checked against Google's official price list and the BazaarLink public catalog: Gemini 3.8 Flash has a standard paid price of US$0.75 / US$3.75 per million input / output tokens; Gemini 3.1 Pro Preview is US$2 / US$12 when the prompt is 200,000 tokens or less. Free tiers are offered per model, so a free allowance for one model does not apply to the whole Gemini API.

Per-Token Prices as of 2026-09-24

Model availability check (2026-09-24): Google lists Gemini 3.1 Flash-Lite as a stable release, and the official prices are the ones shown in this table. BazaarLink currently offers this series under the model ID gemini-3.1-flash-lite-preview. The models you can actually call and their prices are determined by the model catalog. The $0.25 / $1.50 in this table are Google's official stable-release prices. Please check the Google stable model documentation, the Google API changelog, the BazaarLink callable model list and the price / cache catalog. The Gemini 3.1 Pro Preview currently available on BazaarLink is listed on its model page. TWD estimates use the Bank of Taiwan US dollar spot selling rate of NT$31.775 on 2026-09-23. This is a budget conversion, not an actual transaction rate. Simple formula: token count × price per million tokens ÷ 1,000,000 × 31.775. Example: using 1 million input and 1 million output tokens with Gemini 3.8 Flash at the standard prices above comes to about US$4.50, or about NT$143. Exchange-rate source: Bank of Taiwan rate today.

Google's official prices vary by model, context length, and service tier such as Standard, Batch or Flex. Free tiers are only offered for the models Google marks as eligible, and the data-use rules differ between free and paid tiers. The BazaarLink catalog snapshot and official price check date: 2026-09-24. See Google Gemini official pricing, BazaarLink callable model IDs, public price and cache fields, and the Gemini 3.8 Flash model page.

Two Sample First API Requests

The official Gemini REST request uses GEMINI_API_KEY and the generateContent endpoint:

export GEMINI_API_KEY="your Gemini API key"
curl "https://generativelanguage.googleapis.com/v1beta/models/gemini-3.8-flash:generateContent" \
  -H "x-goog-api-key: $GEMINI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"contents":[{"parts":[{"text":"Introduce the Gemini API in one sentence in Traditional Chinese."}]}]}'

Calling BazaarLink with the OpenAI Chat Completions format:

export BAZAARLINK_API_KEY="your BazaarLink key"
curl https://api.bazaarlink.ai/v1/chat/completions \
  -H "Authorization: Bearer $BAZAARLINK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"gemini-3.8-flash","messages":[{"role":"user","content":"Introduce the Gemini API in one sentence in Traditional Chinese."}]}'

For the official steps to apply for a Gemini API key and how to import a project, see the Taiwan Gemini API application guide. For OpenAI and Claude pricing and usage, see the cross-model comparison.

Checked on: 2026-09-24. Free-tier terms and unit prices are taken from the Google official pricing page and rate-limit page as they stood that day. Google changes these often, so verify again before you start. BazaarLink prices are the customer prices shown publicly on the Models page.

People who search for this article fall into two groups: those who want to know whether the Gemini API still has a free quota and how much is left, and those who are budgeting and want to know what each model costs per month in TWD. Google's official pages publish these two things separately, and the free-tier half has recently become hard to look up. This article puts both on one page.

Short answer first:

• The free tier still exists, but only for the Flash family (3.8 / 3.7 / 3.6 / 3.5 Flash, 3.5 and 3.1 Flash-Lite, 2.5 Flash / Flash-Lite). 3.1 Pro and 2.5 Pro have no free tier. Gemma 4 is free.
• The specific RPM / TPM / RPD figures for the free tier have been removed from the official documentation page and can now only be seen on your own AI Studio project page. When people say the quota was "cut" or "shrunk," this is most likely why: it is now shown per project rather than in a public table.
• Content on the free tier may be used by Google to improve its products. Content on the paid tier is not. This is stated on the pricing page, but few people notice it.
• For paid prices within the Flash tier, use 3.8 Flash ($0.75 / $3.75; the official price is marked through 2026-12-31). Its input is half the price of 3.5 Flash, and its output is 58% cheaper.

Don't want to open a GCP project and want to use a company tax ID for invoices when calling Gemini? BazaarLink provides an OpenAI-compatible endpoint, with one key for mainstream models. For the expense-claim process, see the complete guide to expensing AI APIs for Taiwan companies.

Gemini API Free Tier and Key Rules (Official Page, 2026-09-24)

Google sets the free tier per model. It is not one shared allowance across the whole Gemini API. The official pricing page lists Gemini 3.8 Flash and Gemini 3.1 Flash-Lite under the Free Tier; for Gemini 3.1 Pro Preview, the Free Tier is marked Not available. Content on the free tier may be used by Google to improve products; for the paid tier this item is marked No. For formal data and confidentiality requirements, check the policy for the service tier first.

The API Keys page in AI Studio manages projects and keys. Google's documentation says that since 2026-05-28, newly created keys default to authorization keys, and that standard keys will be rejected starting September 2026. For projects still using older keys, log in to AI Studio, check the Key Type, and update according to the official migration documentation. Per-model limits such as RPM / TPM are set at the project level, so follow what AI Studio displays. Do not apply numbers from other accounts.

Source check date: 2026-09-24: Gemini official pricing and free tier, Gemini API key types and migration.

How to Set a Budget Cap so You Don't Get a Surprise Bill

• On the Google side: the Tier itself is a monthly cap (Tier 1 is $250). You can also set budget alerts in Cloud Billing. Note that these are alerts, not automatic shutoffs.
• On the BazaarLink side: each API key can have a spending cap, and once reached, requests are blocked (not just alerted). The free tier is a fixed daily request count. Setting both layers is more reliable than setting just one.

Gemini API Price Comparison (2026-09-24, USD / TWD Estimates)

Prices are calculated by model, token count, and service tier. The table below uses the Standard paid prices per million text tokens. The Gemini 3.1 Pro Preview price is split by whether the prompt exceeds 200,000 tokens.

TWD conversion uses the Bank of Taiwan US dollar spot selling rate of NT$31.775 on 2026-09-23, for budget estimates only. The bank notes that its posted rate does not represent the actual transaction rate. To calculate, first compute the USD amount: input tokens ÷ 1,000,000 × input price + output tokens ÷ 1,000,000 × output price, then multiply by the estimated exchange rate. For Gemini 3.8 Flash with 1 MTok of input and 1 MTok of output, the total is about US$4.50 / NT$143.

Price source check date: 2026-09-24: Google Gemini API pricing, Bank of Taiwan exchange rate, BazaarLink public model IDs and public price catalog. The latter is used to compare live customer prices and cache and promotion fields. Individual model pages: Gemini 3.8 Flash and BazaarLink Gemini 3.1 Pro Preview.

Three Scenario Estimates (Monthly Cost)

The following keeps the workload assumptions from the original article and recalculates them with the 2026-09-24 official Standard prices and the NT$31.775 exchange rate. Other features, storage, and extra tool fees are not included.

Scenario 1: Customer Service Bot (Gemini 3.8 Flash)

500 conversations a day, 3 turns each, with 300 input and 150 output tokens per turn, estimated over 30 days: 13.5 MTok input and 6.75 MTok output. Cost is about US$35.44, or about NT$1,127.

Scenario 2: Document Summarization (Google Official Gemini 3.1 Flash-Lite)

100 documents a day, 6,500 input tokens per document, and 800 output tokens per summary, estimated over 30 days: 19.5 MTok input and 2.4 MTok output. Cost is about US$8.48, or about NT$269.

Scenario 3: Long Document Analysis (Gemini 3.1 Pro Preview)

50 documents a day, 50,000 input tokens per document, and 2,000 output tokens, estimated over 30 days: 75 MTok input and 3 MTok output. When each prompt is under 200K tokens, the cost is about US$186, or about NT$5,910. If a single prompt exceeds 200K, recalculate using the higher official tier.

Actual TWD amounts change with the daily exchange rate and your payment channel. For more on creating a key, see the Taiwan Gemini API application guide. For a side-by-side comparison of OpenAI, Claude and Gemini, see the AI API pricing comparison.

Usage Rebates: How Much Comes Back

BazaarLink's usage rebate is not a percentage discount. It is a fixed milestone rebate: each time the eligible spending in a month crosses a threshold, a fixed amount is paid. Reached thresholds accumulate (not just the highest one), and spending beyond the last threshold earns a final rebate at the tail rate. Rebates go directly into your account balance.

The Gemini series uses the Gemini rebate plan, whose thresholds are the same as the GPT plan:

The rebate rate is not fixed; it varies with the month's spending. It is best when you land exactly on a threshold, and it is diluted between thresholds:

Applying the three scenarios above:

Scenario 2 is a useful reminder: Gemini's lightweight models are cheap enough that they rarely reach a rebate threshold. If your monthly Gemini spending is only a few dollars, the rebate doesn't matter to you, and you should just look at the unit prices.

A few things to know first:

• Settlement uses the UTC calendar month. Rebates for the previous month are credited on the 1st of each month, and are recalculated every month rather than accumulating into the next month.
• The rebate is a balance credit, not a payment item, so it is not refunded and no separate invoice is issued. It also does not expire.
• Free-plan usage, bring-your-own-key (BYOK) usage, and usage covered by subscription credits do not count.
• Rebates go to the wallet that generated the usage: personal usage goes to the personal wallet, and organization usage goes to the organization wallet.
• Each plan is calculated separately, so spending on different plans does not combine toward thresholds.

For the full rules and the list of models currently eligible, see the usage rebate details.

BazaarLink vs. Buying Directly from Google

→ For the full expense-claim process, see: How to expense AI API costs in a Taiwan company? 2026 complete guide

Closing

Want to test quality before deciding? BazaarLink free trial (no credit card required) lets you use the free tier to get your code running, then switch to a paid Gemini-series model. Related pricing articles: Claude TWD cost estimate, GPT-5 TWD cost estimate, DeepSeek TWD cost estimate. For the latest prices, see BazaarLink Models.

TWD billing · Taiwan invoices · leading AI models · OpenAI-compatible API