Google Gemini Developer API pricing (serverless, per 1M tokens, 2026-10-04 list price)

USD per 1M TOKENS, LIST PRICE, observed 2026-10-04, page last updated 2026-10-01 UTC. Gemini 3.8 Flash Standard tier: input $0.75, output $3.75, context-caching reads $0.075, all through December 31 2026 - each of those DOUBLES on January 1 2027 to $1.50 / $7.50 / $0.15. Same model, other tiers: Batch $0.375 in / $1.875 out; Flex $0.375 in / $1.875 out; Priority $1.35 in / $6.75 out. price_usd = Standard input. Explicit storage charge on the page: context-caching STORAGE is $0.50 per 1M tokens PER HOUR through 2026-12-31, doubling to $1.00 from 2027-01-01 - a per-hour recurring charge on cached context, not a one-off write. Grounding add-ons: Google Search and Maps 5,000 free requests per month shared across Gemini 3.x, then $14 per 1,000 requests; embeddings for File search $0.15 per 1M tokens. The identical Standard/Batch/Flex numbers appear on Gemini 3.7 Flash and 3.6 Flash, i.e. the Flash tier is priced as one band regardless of generation PERISHABLE-FIGURE STAMP (added 2026-10-04, uniform across data/supply/ so it is greppable). PRICE CLASS: PUBLISHED LIST PRICE - an asking rate a new customer is quoted, NOT a realised, discounted or negotiated transaction rate. A real bill is at or below this and typically well below it once commitments, credits and committed-use discounts apply. OBSERVED: 2026-10-04 RE-VERIFY BY: 2027-01-04 - the repo's quarterly re-verification bar. After that date this figure is presumed drifted and must be re-checked against its source before it is quoted again. A price without its observation date is noise; this stamp exists so a reader can tell a fresh figure from a stale one without reading prose. UNIT: USD per 1M TOKENS, LIST PRICE, observed 2026-10-04, page last updated 2026-10-01 UTC READING RULE: No further perishable flag is published for this figure. KNOWN FUTURE CHANGE - PUBLISHED, DATED, AND NOW EXHAUSTIVE ACROSS ALL FOUR TIERS, verified against the live page on 2026-10-04 (page footer: 'Last updated 2026-10-01 UTC'): EVERY Gemini 3.x Flash rate is published as a PAIR, 'through December 31, 2026' and 'starting January 1, 2027', and every one of the 2027 values is 2x its 2026 value. Standard: input $0.75 -> $1.50, output $3.75 -> $7.50, context-caching read $0.075 -> $0.15, context-caching STORAGE $0.50 -> $1.00 per 1M tokens per hour. Batch: input $0.375 -> $0.75, output $1.875 -> $3.75, cache read $0.0375 -> $0.075. Flex: IDENTICAL to Batch on every line, same doubling. Priority: input $1.35 -> $2.70, output $6.75 -> $13.50, cache read $0.135 -> $0.27. SO THE PRICE STRUCTURE ITSELF CHANGES, NOT JUST ONE COLUMN: the ratio between Priority and Standard is 1.8x today ($1.35/$0.75) and 1.8x after ($2.70/$1.50) so tier ORDERING is preserved, but Batch and Flex remain at 0.5x Standard before and after, while the Priority/Standard gap on OUTPUT widens from 1.8x to 1.8x - check any cost model that hardcodes a tier multiplier rather than reading the live table. THE IMPORTANT CORRECTION TO THE PREVIOUS REVISION OF THIS RECORD: an earlier pass recorded the doubling as applying to the Standard tier only. Re-read on 2026-10-04, Google publishes the dated 2x pair for Batch, Flex AND Priority as well, so a model that switches tier at 2027-01-01 does not escape the increase. VALIDITY WINDOW OF THE STORED FIGURE: price_usd 0.75 and every other 2026 figure in this record are TRUE ONLY THROUGH 2026-12-31. From 2027-01-01 the stored 0.75 is wrong by exactly 2x and must be replaced, not extrapolated.

lead_time_weeks is null because NO NUMBER APPLIES, and the reason is specific rather than generic: this record prices a metered service term (tokens, egress, storage) rather than hardware capacity, so there is no order-to-delivery interval to report at all. Calling this 'no published lead time' would be wrong in the other direction - nothing was withheld.

no EAR citation: no controlled item is transferred by this transaction; not an EAR citation: provider service terms and end-use policy (the AI Diffusion Rule briefly tiered service access before its 2025-05-13 rescission)

MECHANISM: a token-priced API, so no accelerator crosses a border and the 3A090 hardware licence regime does not attach to the transaction. Export control becomes an END-USER AND TERMS question instead: the provider applies geographic and end-use restrictions to service access, and a customer can still be refused under sanctions or end-use policy wherever the compute sits. A reader targeting a restricted destination should read the provider's terms, not the EAR.

A DATE-BOUND PRICE WITH A PUBLISHED CLIFF IS THE MOST DANGEROUS KIND TO CACHE, AND THIS IS ONE. Every Gemini 3.x Flash figure on this page is explicitly 'through December 31, 2026' with a stated doubling 'starting January 1, 2027'. Any comparison written today is only valid for the remainder of 2026 and a reader in 2027 who trusts a stored $0.75 will be wrong by exactly 2x. Note also that Batch and Flex are priced IDENTICALLY here (both $0.375 in / $1.875 out) whereas at OpenAI Flex and Batch are also equal but at a different multiplier - do not assume the tiers are portable between vendors. Second structural point: this is the only provider in this set that prices cached-context STORAGE explicitly and per hour, which for a large stable prefix can dominate the token bill. Google publishes no GPU hourly rate for its Gemini serving fleet. EXPORT CONTROL (2026-10-04): NOT APPLICABLE IN THE HARDWARE-SHIPMENT SENSE - this record prices a token-based API, so no accelerator is transferred across a border and the ECCN 3A090 hardware licence regime does not attach to the transaction. Recorded null rather than false precisely because a token API is where export control instead becomes an END-USER/TERMS question: the provider applies geographic and end-user restrictions to service access (the AI Diffusion Rule briefly put a tiered country structure on exactly this kind of service before being rescinded on 2025-05-13), and a customer can still be refused under sanctions or end-use policy regardless of where the compute sits. Anyone reading this record for a restricted destination should look at the provider's terms, not at the EAR. Sources for the regime: spl-bis-2025-05-ai-diffusion-rescission, spl-bis-2026-05-d5-guidance. EXPORT CONTROL (2026-10-04): NOT APPLICABLE IN THE HARDWARE-SHIPMENT SENSE - this record prices a token-based API, so no accelerator crosses a border and the ECCN 3A090 hardware licence regime does not attach to the transaction. Recorded NULL rather than false precisely because a token API is where export control instead becomes an END-USER and TERMS question: the provider applies geographic and end-user restrictions to service access (the AI Diffusion Rule briefly put a tiered country structure on exactly this kind of service before being rescinded and unenforced from 2025-05-13), and a customer can still be refused under sanctions or end-use policy regardless of where the compute sits. For a restricted destination, read the provider's terms, not the EAR. Sources: spl-bis-2025-05-ai-diffusion-rescission, spl-bis-2026-05-d5-guidance. LEAD TIME: null - correct and not a gap. A token-priced serverless API is provisioned on request, so there is no lead time to publish; the binding constraint when a model is at capacity is provider-side queuing and any per-account concurrency limit, which this repo records in notes where a vendor publishes one (see the DeepSeek concurrency figures already here). Deliberately NOT recorded: any 'model capacity lead time' estimate, because vendors publish none and the SEO 'GPU lead time' numbers that circulate do not apply to a rented endpoint. Probe log in source spl-negative-no-oem-publishes-lead-time. STALENESS FLAG - RE-VERIFIED AGAINST THE LIVE PAGE 2026-10-04, AND THE CLIFF IS LARGER THAN PREVIOUSLY RECORDED. The stored $0.75 remains correct today and expires in 88 days. The earlier revision of this note recorded a doubling on 2027-01-01 for the Standard tier's input, output, cache-read and cache-storage charges. Re-reading the pricing page on 2026-10-04 shows the same dated doubling published for the Batch, Flex and Priority tiers as well: Standard $0.75/$3.75/$0.075 -> $1.50/$7.50/$0.15; Batch $0.375/$1.875 -> $0.75/$3.75; Flex identical to Batch; Priority $1.35/$6.75/$0.135 -> $2.70/$13.50/$0.27; and cache STORAGE $0.50 -> $1.00 per 1M tokens per hour on every tier. THAT IS THE WHOLE PAGE. There is no Gemini 3.x Flash rate that does not double, so there is no tier to migrate to in order to hold cost flat - a reader planning across 2027-01-01 should apply the published 2x rather than assume a cheaper tier appears. WHY THIS IS THE RECORD WITH THE SHORTEST VALIDITY WINDOW IN data/supply/: it is not merely old, it is EXPIRED-BY-DATE. Every other priced record in this directory is merely 'possibly drifted'; this one is 'definitely wrong' from a known calendar date. That is a better property than stability, because it is actionable: a consumer of this record can compute the correct future price today, whereas a stale record can only be re-fetched after the fact. The generic PERISHABLE-FIGURE STAMP elsewhere in this record carries a quarterly re-verify date of 2027-01-04, which for this record is nearly a month LATE relative to its own published cliff - the cliff governs here, not the bar. WHAT ELSE ON THE PAGE IS NOT A PRICE, restated to stop it being quoted as one: '5,000 free search requests per month (shared across all Gemini 3.x models), then $14 per 1,000 requests' is a request-rate allowance plus an overage rate, not a token rate; the 'Google AI Studio usage is free of charge in all available regions' line applies to AI Studio only and says nothing about paid API usage; and the pricing-for-agents section states that agentic loops are billed 'at standard Gemini list rates, including input, output, and intermediate input / reasoning tokens' - i.e. reasoning tokens are NOT free and are charged as output. That last point is an inference-cost trap independent of the date cliff. STILL UNVERIFIED: the page's own caveat says 'Prices may differ from the prices listed here and the prices offered on Gemini Enterprise Agent Platform', and Enterprise discounts are volume-based and quote-only. No Enterprise price was obtained.