Google

google.dev · Compute, Inference, Storage · 7 records · JSON

Gemini API code execution Computecaveatemail?card?Rper-object-cap
Google Gemini API Computecaveatemailno cardDrequest-rate
Google Gemini API Inferencecaveatemailno cardDrequest-rate
Gemini File Search stores (managed RAG) Storagefails: quotaemailno cardRper-object-cap
gemini file search stores as bulk storage Storagefails: quotaemail?card?R
gemini files api (ai studio) as an archive Storagefails: quotaemail?card?R
Google Gemini Files API (AI Studio) Storagefails: quotaemailno cardRper-object-cap

Gemini API code execution Compute

caveatemail?card?Rper-object-cap

ai.google.dev/gemini-api/docs/code-execution · container and VM sandbox · compute/gemini-code-execution · JSON · compare

Free tier	not published unit none; metered on wall-clock
Price	none stated
Per-object cap	not stated
Retention	not stated
Egress	not stated
Identity	email (placeholder) not published
Limits	per-object-cap the maximum input file size is bounded by the model's token window, not by a code-execution quota source excluded as too-big
Binding	per-object-cap
Interface	unknown
Tooling	not stated
Composes with	not stated
Reliability	no reliability event recorded
Evidence rung	reported the code-execution page was read on 2026-10-22.
Checked	2026-10-23 (live); stale on 2026-11-06

Evidence (2)

date	source	what it proved
2026-10-02	ai.google.dev/gemini-api/docs/code-execution	not published
2026-10-22	ai.google.dev/gemini-api/docs/code-execution	the code-execution documentation was read on 2026-10-22 and describes the tool and its limitations; it states no free allowance of its own, so the boundary is the model's token window

More fields (13)

{
 "talaria_m10h_usd": null,
 "talaria_m24h_usd": null,
 "nemo_m10h_usd": null,
 "keep_rate": null,
 "keep_rate_basis": "no field in the source",
 "billing_model": "unclassified",
 "shape": null,
 "session_cap": {
  "wall_clock_min": null,
  "idle_reap_s": null,
  "restart_required": false,
  "note": "the page states no execution clock; the sandbox lives for the call"
 },
 "persistent_storage": null,
 "network_out": null,
 "price_note": null,
 "free_tier_kind": "allowance",
 "net": null
}

Unknown or withheld values (17)

$.chaining not-stated
$.egress.metered egress-unknown
$.egress.terms not-stated
$.evidence[0].proved process
$.extras.price_note process
$.free_tier.notes withheld
$.free_tier.stated process
$.identity_cost.detail defaulted
$.identity_cost.kind identity-defaulted
$.interface interface-unknown
$.interface_detail withheld
$.per_object_cap.notes process
$.per_object_cap.stated not-stated
$.reliability_note none-observed
$.retention.notes not-extracted
$.retention.stated not-stated
$.tooling not-stated

Google Gemini API Compute

caveatemailno cardDrequest-rate

ai.google.dev/gemini-api/docs/pricing · inference API · compute/google-gemini-api · JSON · compare

Free tier	The free tier still exists and the pricing page still prints 'Free of charge' against input, output and caching for Gemini 3.8 Flash, 2.5 Flash-Lite and a dozen older models, but Google has stopped publishing how much the free tier carries. The rate-limits page now says only 'Rate limits depend on a variety of factors (such as your usage tier) and can be viewed in Google AI Studio', and the per-model free table the page carried through 2025 has been removed. The free columns that remain describe what is free, never how much. A Google AI Developers Forum post dated 2026-09-03 quotes 20 RPD on Gemini 3.8 Flash against 500 RPD on Flash-Lite. unit none; metered on wall-clock; recurring-daily-request-allowance
Price	none stated
Per-object cap	not applicable
Retention	not applicable
Egress	not stated
Identity	email not published
Limits	request-rate requests per day binds, not tokens: at a reported 20 RPD on Gemini 3.8 Flash the monthly ceiling is 600 requests, so a 1,000-item job at three requests an item never completes on one free project. source requests per day binds, not tokens: at a reported 20 RPD on Gemini 3.8 Flash the monthly ceiling is 600 requests, so a 1,000-item job at three requests an item never completes on one free project.
Binding	request-rate requests per day binds, not tokens: at a reported 20 RPD on Gemini 3.8 Flash the monthly ceiling is 600 requests, so a 1,000-item job at three requests an item never completes on one free project.
Interface	rest inferred from the tooling the provider documents; not probed
Tooling	not stated
Composes with	the free allowance is not the reason to use it: the Batch API at 100 concurrent requests and a 2GB input file is a paid-tier feature and the free-tier Batch rows read 'Not available', so a real batch needs the billing account the free tier exists to avoid.
Reliability	no reliability event recorded
Notes	Arithmetic, labelled estimate: 20 RPD x 30 days = 600 requests a month per model. The gap is the finding: a reader who trusts the word 'generous' on the pricing page and not the forum will design a batch that cannot run. Not measured: limits appear only in AI Studio after an account exists, so the 20 RPD figure rests on a forum report, and the 2025 archive's 250 RPD for 2.5 Flash disagrees with it by an order of magnitude. The corpus already carries google-gemini-api as qualifies-with-caveat on request-rate; this record adds the capacity shape and the fact that the number is no longer public.
Evidence rung	documented
Checked	2026-10-22 (live); stale on 2026-11-21
Parsed numbers	`$.limits[0].value` limit:request-rate 600.0 requests

Evidence (4)

date	source	what it proved
2026-10-22	ai.google.dev/gemini-api/docs/pricing	Free Tier column reads 'Free of charge' for input, output and context caching on Gemini 3.8 Flash, while the Batch, Flex and Grounding rows read 'Not available', quoted
2026-10-22	ai.google.dev/gemini-api/docs/rate-limits	'Rate limits depend on a variety of factors (such as your usage tier) and can be viewed in Google AI Studio'; the page carries no free-tier RPM/RPD/TPM table, quoted
2026-09-03	discuss.ai.google.dev/t/gemini-3-8-flash-free-tier-20-rpd-is	developer post states 'the free tier limit of 20 RPD is very restrictive' and contrasts 'Flash-Lite's 500 RPD'; reported, not a vendor figure
2026-10-22	web.archive.org/web/20251201061102/https://ai.google.dev/gem	the archived page still carried a Free Tier table with Gemini 2.5 Flash at 10 RPM, 250,000 TPM and 250 RPD, which the live page no longer carries [re-read 2026-10-22]

More fields (1)

{
 "wave2": {
  "class": "capacity-shape",
  "free_tier_kind": "recurring-daily-request-allowance",
  "shape": {
   "vcpu": null,
   "ram_gib": null,
   "gpu_class": null,
   "gpu_count": null
  }
 }
}

Unknown or withheld values (9)

$.egress.metered egress-unknown
$.egress.terms not-stated
$.identity_cost.detail process
$.per_object_cap.notes process
$.per_object_cap.stated not-applicable
$.reliability_note none-observed
$.retention.notes process
$.retention.stated not-applicable
$.tooling not-stated

Google Gemini API Inference

caveatemailno cardDrequest-rate

ai.google.dev/gemini-api/docs/pricing · inference endpoint, free tier · inference/google-gemini-api · JSON · compare

Free tier	gemini-3.8-flash is 'Free of charge' for input and output on the Free Tier unit tokens; metered on request-count
Price	none stated
Per-object cap	not applicable: an inference endpoint stores nothing for the caller
Retention	not applicable
Egress	the response is the payload; no separate egress term is published
Identity	email a Google account and an AI Studio project; no card for the free tier
Limits	request-rate RPM, input TPM and RPD per project; RPD resets at midnight Pacific; the values are visible only in AI Studio source
Binding	request-rate
Interface	rest endpoint: https://generativelanguage.googleapis.com/v1beta/
Tooling	any OpenAI-compatible client; curl against POST {base}/chat/completions
Composes with	pairs with an agent as the email-gated fallback; the scarce link is the per-project allowance behind AI Studio, which no page states
Reliability	free-tier content is used to improve products; the paid tier is the one the vendor describes for production
Notes	The binding fact is that the free tier publishes no number. A reader planning against '15 RPM, 1,500 RPD' is planning against an aggregator.
Evidence rung	documented the price and the gating are documented. The numeric limits are not public, so no number is carried and any catalogue that prints one is quoting a third party. verified_at is the instant its evidence URLs were fetched on 2026-10-22, read from the fetch log, not typed.
Checked	2026-10-23 (live); stale on 2026-11-22

Evidence (1)

date	source	what it proved
2026-10-22	ai.google.dev/gemini-api/docs/rate-limits	limits are applied per project and not per API key; RPD resets at midnight Pacific; 'Specified rate limits are not guaranteed and actual capacity may vary'; the per-model numbers are served only at aistudio.google.com/rate-limit

Unknown or withheld values (1)

$.retention.stated not-applicable

Gemini File Search stores (managed RAG) Storage

fails: quotaemailno cardRper-object-cap

ai.google.dev/gemini-api/docs/file-search · Managed vector store · storage/gemini-file-search-stores · JSON · compare

Free tier	Free tier 1 GB of store; Tier 1 10 GB; Tier 2 100 GB; Tier 3 1 TB. 100 MB per document. unit GB; metered on stored-bytes
Price	none stated
Per-object cap	per-object-cap 100 MB (100000000 bytes)
Retention	Raw files are deleted after 48 hours; only chunked embeddings persist, and there is no documented route to reassemble originals (2 days)
Egress	Storage free, query-time embeddings free. You pay only indexing embeddings ($0.15 / 1M tokens) and context tokens on retrieval.
Identity	email email
Limits	retention-clamped deleted after 48 hours source per-object-cap per-object-cap 100 MB source per-object-cap 100 MB. 'Recommendation: Limit the size of each File Search store to under 20 GB'. No TTL on embeddings — they persist until manually deleted or the model is deprecated.
Binding	per-object-cap the per-object cap sets the shard size and therefore the object count; it bounds the total more tightly than any headline quota
Interface	rest google-genai `client.file_search_stores.create` / `uploadToFileSearchStore` / `documents.list`
Tooling	pip install google-genai; REST fileSearchStores.documents
Composes with	Multiple stores per project, but the project-wide total is the cap. 'File Search store size is computed on the backend … typically approximately 3 times the size of your input data' — so 1 TB of Tier 3 store ≈ 340 GB of input. On free tier, 1 GB store ≈ 333 MB of input.
Reliability	Embeddings are not bytes. Raw files are deleted after 48 hours; only chunked embeddings persist, and there is no documented route to reassemble originals. The 3x expansion factor is the trap — the marketing number understates the cost of the input by 3x. Free tier at 1 GB is below the brief's floor by a factor of 500.
Notes	Free tier 1 GB of store; Tier 1 10 GB; Tier 2 100 GB; Tier 3 1 TB. 100 MB per document.
Evidence rung	reported One page read moves it to documented.
Checked	2026-10-21 (inherited); stale on 2026-11-04
Parsed numbers	`$.limits[0].value` limit:retention-clamped 48.0 hours `$.limits[1].value` limit:per-object-cap 100.0 MB

Evidence (4)

date	source	what it proved
2026-10-21	ai.google.dev/gemini-api/docs/file-search	'Total size of project File Search stores (based on user tier): Free: 1 GB; Tier 1: 10 GB; Tier 2: 100 GB; Tier 3: 1 TB'; 'Maximum file size / per document limit: 100 MB'
2026-10-21	ai.google.dev/gemini-api/docs/file-search	'There is no Time To Live (TTL) for embeddings; they persist until manually deleted … Files, however, are deleted after 48 hours.'
2026-10-21	ai.google.dev/gemini-api/docs/file-search	'This is typically approximately 3 times the size of your input data'; 'Storage is free of charge.'
2026-10-21	blog.google/innovation-and-ai/technology/developers-tools/fi	'we're making storage and embedding generation at query time free of charge … $0.15 per 1 million tokens'

More fields (1)

{
 "dead_end": {
  "failure_class": "quota",
  "why": "Free tier 1 GB of store; Tier 1 10 GB; Tier 2 100 GB; Tier 3 1 TB. 100 MB per document.",
  "retry_when": null,
  "first_recorded": "2026-10-21"
 }
}

Unknown or withheld values (4)

$.extras.dead_end.retry_when process
$.free_tier.notes process
$.per_object_cap.notes not-extracted
$.retention.notes not-extracted

gemini file search stores as bulk storage Storage

fails: quotaemail?card?R

ai.google.dev/gemini-api/docs/file-search · rejected-provider · storage/gemini-file-search-stores-as-bulk-storage · JSON · compare

Free tier	Free tier is 1 GB of store, and store size runs ~3x input size, so real input capacity is ~333 MB. Only embeddings persist; raw files are deleted after 48h. unit none; metered on none
Price	none stated
Per-object cap	not recorded
Retention	not recorded
Egress	not recorded
Identity	email (placeholder) not published
Binding	none no constraint of a named kind was recorded; the verdict rests on the quota sentence rather than on a named limit
Interface	unknown
Tooling	not recorded
Composes with	not recorded
Reliability	Free tier is 1 GB of store, and store size runs ~3x input size, so real input capacity is ~333 MB. Only embeddings persist; raw files are deleted after 48h.
Notes	Free tier is 1 GB of store, and store size runs ~3x input size, so real input capacity is ~333 MB. Only embeddings persist; raw files are deleted after 48h.
Evidence rung	reported
Checked	2026-10-21 (inherited); stale on 2026-11-04

Evidence (1)

date	source	what it proved
2026-10-21	ai.google.dev/gemini-api/docs/file-search	Free tier is 1 GB of store, and store size runs ~3x input size, so real input capacity is ~333 MB. Only embeddings persist; raw files are deleted after 48h.

More fields (1)

{
 "dead_end": {
  "failure_class": "retention",
  "why": "Free tier is 1 GB of store, and store size runs ~3x input size, so real input capacity is ~333 MB. Only embeddings persist; raw files are deleted after 48h.",
  "retry_when": null,
  "first_recorded": "2026-10-21"
 }
}

Unknown or withheld values (15)

$.chaining not-recorded
$.confidence_note withheld
$.egress.terms not-recorded
$.extras.dead_end.retry_when process
$.free_tier.notes process
$.identity_cost.detail defaulted
$.identity_cost.kind identity-defaulted
$.interface interface-unknown
$.interface_detail withheld
$.limits_note process
$.per_object_cap.notes not-recorded
$.per_object_cap.stated not-recorded
$.retention.notes process
$.retention.stated not-recorded
$.tooling not-recorded

gemini files api (ai studio) as an archive Storage

fails: quotaemail?card?R

ai.google.dev/gemini-api/docs/files · rejected-provider · storage/gemini-files-api-as-an-archive · JSON · compare

Free tier	48-hour retention, clamped server-side, plus 'you can't download user-uploaded files'. 20 GB/project is moot. unit none; metered on none
Price	none stated
Per-object cap	not recorded
Retention	not recorded
Egress	not recorded
Identity	email (placeholder) not published
Binding	none no constraint of a named kind was recorded; the verdict rests on the quota sentence rather than on a named limit
Interface	unknown
Tooling	not recorded
Composes with	not recorded
Reliability	48-hour retention, clamped server-side, plus 'you can't download user-uploaded files'. 20 GB/project is moot.
Notes	48-hour retention, clamped server-side, plus 'you can't download user-uploaded files'. 20 GB/project is moot.
Evidence rung	reported
Checked	2026-10-21 (inherited); stale on 2026-11-04

Evidence (1)

date	source	what it proved
2026-10-21	ai.google.dev/gemini-api/docs/files	48-hour retention, clamped server-side, plus 'you can't download user-uploaded files'. 20 GB/project is moot.

More fields (1)

{
 "dead_end": {
  "failure_class": "retention",
  "why": "48-hour retention, clamped server-side, plus 'you can't download user-uploaded files'. 20 GB/project is moot.",
  "retry_when": null,
  "first_recorded": "2026-10-21"
 }
}

Unknown or withheld values (15)

$.chaining not-recorded
$.confidence_note withheld
$.egress.terms not-recorded
$.extras.dead_end.retry_when process
$.free_tier.notes process
$.identity_cost.detail defaulted
$.identity_cost.kind identity-defaulted
$.interface interface-unknown
$.interface_detail withheld
$.limits_note process
$.per_object_cap.notes not-recorded
$.per_object_cap.stated not-recorded
$.retention.notes process
$.retention.stated not-recorded
$.tooling not-recorded

Google Gemini Files API (AI Studio) Storage

fails: quotaemailno cardRper-object-cap

ai.google.dev/gemini-api/docs/files · LLM provider /v1/files · storage/google-gemini-files-api · JSON · compare

Free tier	20 GB per project; 2 GB per file; 48-hour retention unit GB; metered on unstated
Price	none stated
Per-object cap	2 GB per file (2000000000 bytes)
Retention	20 GB per project; 2 GB per file; 48-hour retention (2 days)
Egress	Free — 'The Files API is available at no cost in all regions where the Gemini API is available.'
Identity	email email
Limits	per-object-cap 2 GB per file source retention-clamped Retention-clamped hard at 48 hours source per-object-cap 2 GB. Retention-clamped hard at 48 hours — you cannot request longer. PDFs additionally capped at 50 MB / 1,000 pages for processing. Inline path is 100 MB total request.
Binding	per-object-cap the per-object cap sets the shard size and therefore the object count; it bounds the total more tightly than any headline quota
Interface	rest REST via google-genai SDK `client.files.upload`; generativelanguage.googleapis.com/v1beta
Tooling	pip install google-genai; curl multipart to /upload/v1beta/files
Composes with	20 GB/project, and Google Cloud projects are the multiplicative axis. But 48-hour retention means capacity, not storage.
Reliability	Double fail, and the docs say it in one sentence: 'Files are stored for 48 hours. During that time, you can use the API to get metadata about the files. However, you can't download user-uploaded files.' This is the brief's `retention-clamped` class plus a read gate. Only model-generated files (video) are downloadable.
Notes	20 GB per project; 2 GB per file; 48-hour retention
Evidence rung	reported One page read moves it to documented.
Checked	2026-10-21 (inherited); stale on 2026-11-04
Parsed numbers	`$.limits[0].value` limit:per-object-cap 2.0 GB `$.limits[1].value` limit:retention-clamped 48.0 hours

Evidence (4)

date	source	what it proved
2026-10-21	ai.google.dev/gemini-api/docs/files	'The Files API lets you store up to 20 GB of files per project, with a per-file maximum size of 2 GB. Files are stored for 48 hours.'; 'you can't download user-uploaded files'
2026-10-21	ai.google.dev/gemini-api/docs/files	'Files are automatically deleted after 48 hours'
2026-10-21	github.com/google-gemini/cookbook/blob/main/quickstarts/File	'The File API accepts files under 2GB in size and can store up to 20GB of files per project. Files last for 2 days and cannot be downloaded from the API.'
2026-10-21	discuss.ai.google.dev/t/uploaded-files-missing-after-a-few-d	'The Files API storage has an expiration time of 48 hrs and is restricted to 20GB total storage'

More fields (1)

{
 "dead_end": {
  "failure_class": "retention",
  "why": "20 GB per project; 2 GB per file; 48-hour retention",
  "retry_when": null,
  "first_recorded": "2026-10-21"
 }
}

Unknown or withheld values (4)

$.extras.dead_end.retry_when process
$.free_tier.notes process
$.per_object_cap.notes not-extracted
$.retention.notes not-extracted