As checked 2026-10-25: Google’s Gemini Developer API pricing table distinguishes free access, paid standard, batch, flex and priority tiers, with model-specific availability. This is a dated economics guide, not a permanent price guarantee or a personal hands-on benchmark.
Scope: global English reference; USD list prices where shown; local eligibility, currency, taxes, negotiated terms and model release dates require separate confirmation. Dynamic source-page prices have a check date, not a verified initial effective date.
Pricing depends on exact model and route
Google’s Gemini Developer API pricing table distinguishes free access, paid standard, batch, flex and priority tiers, with model-specific availability. In the October 25, 2026 observation, the Gemini 3.1 Flash-Lite paid standard text/image/video input rate was $0.25 per million tokens, with $1.50 output and $0.025 cached input. Its batch input and output were shown at $0.125 and $0.75 for those modalities. Audio input has a different rate. These are first-party live-page observations; no original rate-effective date is published in this extraction.
A Pro prompt boundary changes both sides
For gemini-3.1-pro-preview, Google’s standard table splits prompts up to 200,000 tokens from those above that length. The input list rates are $2 and $4 per million respectively; the corresponding output rates are $12 and $18 per million including thinking. The batch rows also split at the same boundary. Do not combine the short-context input rate with the long-context output rate, and do not pretend this preview ID is identical to a later stable deployment.
Caching is not just a cheaper read
The page shows a separate storage rental for explicit cache retention, such as $4.50 per million token-hours for the cited Pro Preview standard tier. If an app stores a one-million-token cache for six billed hours at that listed rate, storage alone is 6 × $4.50 = $27, before any cached reads, cache creation or output. This is an illustrative use of the official rate, not a claim that an application should create such a cache. Resolve actual retention and billing precision from the API response and documentation.
Free tier and tool add-ons
Google’s billing guide describes separate free and paid access and qualification for different billing tiers; free-tier availability is not an unlimited production SLA. Tool use and grounding can create other charges or allowances distinct from text inference. A user running a WordPress editorial service should check the account’s own AI Studio quota view instead of inferring entitlement from another region. Batch discount claims are conditional on the specific endpoint and listed SKU.
Cost-control decisions
Record model ID, paid/free channel, media input type, processing tier, prompt-length bracket, cached-token storage time and grounding operations. Use measured usage rather than counting visible English words. Compare a small, cheap model for high-volume transformations with a higher-capability model for escalations, but quality and suitability belong to a separate evaluation; this article supports arithmetic, not a vendor ranking.
A mixed-mode budget example
Imagine an initial classification pass on Flash-Lite followed by escalations on Pro Preview for difficult documents. Count how many records enter each stage and measure each stage’s token use independently. Do not apply Flash-Lite’s rate to Pro escalations or reuse a short-context Pro rate when the source packet exceeds the documented threshold. If a Gemini call includes video, image or audio inputs, the modality can change the input meter. The scenario should also include search-grounding and cache-storage components where used.
Quota and availability verification
Before a launch, validate API billing-account status, the exact preview model identifier and per-account quotas in the authorized dashboard. A model may be described on a global public pricing page while access is restricted to eligible projects or regions. Some historical Gemini variants carry shutdown warnings; archived articles must not imply that a retired ID remained usable. Export the observed tariff and keep its checked timestamp next to the quote, but never backdate that live observation to a guessed release date.
Decision checklist
- Identify the specific plan or API endpoint, model version, region and currency before pricing a job.
- Measure input, output, caching, generated media and failed-job units separately.
- Check the official pricing page again before procurement, especially if the SKU has no effective-date notice.
- Keep source evidence and a timestamped price ID; do not treat unknowns as zero.
Pricing evidence boundary
This cost discussion concerns Google. Reproduce a quote only with its exact model or plan identifier, chargeable unit, billing region, processing mode and timestamp; a changing vendor rate card is not a permanent price guarantee. If any one of these dimensions is missing, keep the associated rate and worked scenario under editorial review instead of filling the gap with an estimate.
Sources & Further Reading
- Gemini Developer API shows tier/model/context/modality-specific rates and cache storage — official first-party source, accessed 2026-10-25; original effective date not specified in source.
- Gemini free/paid billing tiers and quota classification — official first-party source, accessed 2026-10-25; original effective date not specified in source.
- Cloud Gemini and tools are billed on cloud SKU/region with separate charges — official first-party source, accessed 2026-10-25; original effective date not specified in source.