We use cookies for analytics and personalized ads. We use cookies for basic analytics and, where we show ads, to personalize them. Privacy PolicyRead our Privacy Policy.
Estimate token consumption and API cost across Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.1 Pro Preview. Calculate multimodal overheads for images, audio, video, PDFs, and prompt context caching.
In short: pick a Gemini tier, run the calculator below to price your requests, then use the usage guide to find your API key, check free-tier limits, and see where else Gemini shows up (Excel, Cursor, Nano Banana, TTS) before you build on it.
Count Tokens LocallyGemini's API pricing isn't one number — it changes by model, by whether your input is text or multimodal, by how much of it comes from cache, and by which service tier you run on. This calculator gives you a fast, model-by-model estimate of standard text token costs so you can compare Gemini's Flash, Flash-Lite, and Pro tiers before committing to one in production. Below the calculator, we break down what a text-only estimate can't show you — image, audio, video, and PDF tokens, thinking and tool-call usage, context caching, and how Batch, Flex, and Priority processing change the final bill.
Three tiers, in order of how much reasoning they can do and what they cost for it.
Built for high-volume jobs
1,048K / 65K max context
Multimodal grounding support
1,048K / 65K max context
Preview-tier reasoning model
1,048K / 65K max context
Adjust your token input parameters or choose a pre-configured enterprise scenario.
Comparing against Gemini 3.5 Flash: Switching background JSON extraction or RAG pipelines to Flash-Lite reduces costs by approx 76%. Escalating complex tasks to 3.1 Pro Preview adds a 33% cost premium—reserve Pro exclusively for difficult multi-step reasoning.
| Model | Context | Input / 1M | Cache / 1M | Output / 1M | Per Call | Monthly Estimate |
|---|---|---|---|---|---|---|
| Gemini 3.1 Pro Preview | 1,048K / 65K | $2.00 | $0.200 | $12.00 | $0.0190 | $475.00 |
| Gemini 3.5 Flash | 1,048K / 65K | $1.50 | $0.150 | $9.00 | $0.0143 | $356.25 |
| Gemini 2.5 Pro | 1,048K / 65K | $1.25 | $0.125 | $10.00 | $0.0139 | $346.88 |
| Gemini 3.6 Flash | 1,048K / 65K | $0.75 | $0.075 | $3.75 | $0.0065 | $163.13 |
| Gemini 3.7 Flash | 1,048K / 65K | $0.75 | $0.075 | $3.75 | $0.0065 | $163.13 |
| Gemini 3.8 Flash | 1,048K / 65K | $0.75 | $0.075 | $3.75 | $0.0065 | $163.13 |
| Gemini 3.5 Flash-Lite | 1,048K / 65K | $0.30 | $0.030 | $2.50 | $0.0034 | $85.25 |
| Gemini 2.5 Flash | 1,048K / 65K | $0.30 | $0.030 | $2.50 | $0.0034 | $85.25 |
| Gemini 3.1 Flash-Lite | 1,048K / 65K | $0.25 | $0.025 | $1.50 | $0.0024 | $59.38 |
| Gemini 2.5 Flash-Lite | 1,048K / 65K | $0.10 | $0.010 | $0.40 | $0.0008 | $19.75 |
Every estimate on this page runs the same formula shown live above: input tokens, cached-read tokens, and output tokens are each multiplied by that model's published per-token rate, summed, then divided by 1,000,000 and multiplied by your monthly call volume. Nothing is randomized or hidden — the full math is visible in the "Exact Billing Formula" box.
This is a text-token estimate. It does not calculate image, audio, video, or PDF tokens (see the Multimodal section below) and it does not include thinking-token or tool-call overhead, which only appear in the API response's own usage metadata after a real call.
Per-token rates are cross-checked against Google's official Gemini API pricing page and updated whenever Google changes them. Rates on this page were last verified 2026-09-29.
Reviewed by the Apex Tech Journal editorial team. Spotted a rate that's out of date? Tell us and we'll fix it — pricing pages like this one are only useful if the numbers are trustworthy.
A rate card tells you the price per million tokens. It doesn't tell you what to enter, which API surface you're even billed through, or whether "Gemini" in your invoice means the API or a completely separate consumer subscription. Here's the part the numbers alone don't cover.
Google AI Studio is the fastest way to get an API key and includes a free tier with rate limits (see the FAQ below). For production workloads, higher throughput, data-residency requirements, or enterprise billing, Gemini is also available through Vertex AI on Google Cloud — same underlying models, different console, billing, and quota system. If you're testing an idea, start in AI Studio; if you're shipping inside an existing GCP setup, use Vertex.
Google publishes an OpenAI-compatibility endpoint for Gemini, so code already written against the OpenAI SDK can often point at Gemini with only the base URL and API key changed. It's a convenience layer, not a pricing change — calls still bill at Gemini's own per-token rates shown in the calculator above, not OpenAI's.
Every Gemini text request ultimately calls the same underlying method. Once it's live, its response includes a usage field you can feed back into the calculator above to replace your estimate with a real number.
pip install google-genai
from google import genai
client = genai.Client() # reads GEMINI_API_KEY from your environment
response = client.models.generate_content(
model="gemini-3.5-flash",
contents="Hello",
)
print(response.usage_metadata)
# -> prompt_token_count, cached_content_token_count, candidates_token_countThese are two different products with two different bills. Gemini Advanced is Google's consumer chat subscription, bundled into a Google One AI plan with its own usage limits — asking "is Gemini Pro worth it" as a subscriber is really a question about that plan. Calling the Gemini Pro model from your own code through the API — what this calculator prices — is billed per token with no subscription involved. If you're building a product or script, you want the API and the numbers above; if you're deciding whether to pay for the chat app, that's a separate call.
Gemini is embedded directly into Docs, Sheets, and Gmail as part of a Workspace add-on, and third-party plugins bring similar Gemini-assisted formulas and summaries into Excel. That usage is billed through the Workspace/Google One subscription, not the per-token API.
Editors and CLIs like Cursor and OpenCode let you plug in a Gemini API key as the model backing an AI coding assistant. When you do, usage is billed exactly like any other API call — run your expected coding-session token counts through the calculator above.
Google's image-generation model, nicknamed Nano Banana, and Gemini's text-to-speech models are billed separately from text tokens — typically per image or per character/second of audio rather than per input token — so this calculator's text-token numbers won't cover an image- or voice-heavy workload. Budget those separately using the official pricing page linked below.
Google's AI Mode search experience runs on Gemini models under the hood, but it's a consumer search feature with no exposed per-token billing — it isn't something you call or budget through this calculator.
Google Gemini treats all media natively within its transformer context. Here is how non-text files translate into token bills:
Standard images are tiled and scaled. A regular high-res image consumes approximately 258 to 300 tokens depending on resolution and aspect ratio.
Audio is sampled continuously. Gemini processes audio at roughly 32 tokens per second (~1,920 tokens per minute of recorded audio).
Video is sampled into image frames (typically 1 frame per second) plus its underlying audio track.
PDFs are processed as rendered visual pages combined with extracted text layers. Each PDF page consumes roughly 258 tokens plus text tokens.
Already estimated your Gemini cost above? See how it stacks up against the other major providers.
Largest context window (1M+ tokens), native multimodal input, strong price-to-throughput on Flash tiers.
Broadest tool/plugin ecosystem; competitive on mid-tier reasoning cost.
Strong long-form reasoning and coding; caching model is similar to Gemini's.
Lowest headline per-token cost among major providers; smaller ecosystem.
The part a plain rate card doesn't show you
Most people can guess a prompt is "long," but nobody can eyeball what 50,000 monthly requests actually cost across Flash vs. Pro — this does that math in one step.
Instead of tab-hopping between Google's pricing page and a spreadsheet, see Flash-Lite, Flash, and Pro costs side by side for the exact same workload.
The moment your request includes an image, PDF, or audio clip, this page tells you so you don't walk away with a silently wrong number.
Quoting a client for an AI feature works better with a defensible monthly estimate than a guess based on the rate card alone.
Four things worth checking before you scale a workload up.
If your application passes a 5,000-token system rulebook or JSON schema on every request, storing it in Gemini's context cache gets you an 80-90% discount on those repeated input tokens for a small monthly storage fee.
Pro is easy to reach for by default, but Flash-Lite handles simple parsing, log analysis, and categorization at $0.10 per million tokens — worth trying before assuming you need the bigger model.
Overnight PDF indexing, transcript enrichment, and other jobs that don't need an immediate response can run through the Batch API for a flat 50% discount on standard rates.
Output tokens cost up to 10x more than input tokens, so a tight max-output setting or a structured JSON schema is often the cheapest fix available.