Replicate

replicate.com · Compute, Storage · 3 records · JSON

Replicate Computepaidemail?card?Drequest-rate
Replicate (r8.im + Files API) Storagefails: quotainstitutionno cardRper-object-capdisagrees
replicate (model hosting) Storagefails: quotaemail?card?Rdisagrees

Replicate Compute

paidemail?card?Drequest-rate

replicate.com/pricing · GPU cloud · compute/replicate · JSON · compare

Free tier	no free tier: hardware bills per second from $0.09 an hour for small CPU to $5.49 an hour for an H100, and a fast-booting fine-tune is billed only while it processes a request unit none; metered on none
Price	none stated
Per-object cap	not stated
Retention	not stated
Egress	not stated
Identity	email (placeholder) not published
Limits	request-rate per-second billing from $0.09 an hour for CPU Small to $40.32 an hour for eight A100s source retention-clamped a fast-booting fine-tune bills only while it is processing; other private models bill for idle time source per-object-cap multi-GPU A100 and H100 capacity beyond two GPUs requires a committed-spend contract source not stated
Binding	request-rate
Interface	unknown
Tooling	not stated
Composes with	not stated
Reliability	no reliability event recorded
Evidence rung	documented the pricing page was read on 2026-10-22 and publishes the full hardware ladder with per-second rates and the idle-billing rule. There is no free tier on it, so the verdict moves to paid-only
Checked	2026-10-23 (live); stale on 2026-11-22

Evidence (2)

date	source	what it proved
2026-10-02	replicate.com/pricing	not published
2026-10-22	replicate.com/pricing	'You'll only be billed for the time the model is active and processing your requests, so you won't pay for idle time' with the hardware ladder: 'CPU (Small) $0.09/hr', 'Nvidia T4 $0.81/hr', 'Nvidia L40S $3.51/hr', 'Nvidia A100 (80GB) $5.04/hr', 'Nvidia H100 $5.49/hr', quoted

More fields (14)

{
 "talaria_m10h_usd": 108.0,
 "talaria_m24h_usd": 259.2,
 "nemo_m10h_usd": null,
 "keep_rate": 1.0,
 "keep_rate_basis": null,
 "billing_model": "unclassified",
 "shape": {
  "vcpu": 4,
  "ram": 8,
  "disk": null,
  "gpu_class": null,
  "gpu_count": 1,
  "note": "CPU Small is 1x CPU and 2 GB at $0.09; an H100 is 13x CPU, 144 GB of RAM and 80 GB of VRAM at $5.49"
 },
 "session_cap": {
  "wall_clock_min": null,
  "idle_reap_s": null,
  "restart_required": false,
  "note": "a fast-booting fine-tune stops billing between requests; the page contrasts this with models that bill for idle time"
 },
 "persistent_storage": null,
 "network_out": null,
 "price_note": null,
 "free_tier_kind": "none",
 "ps": null,
 "net": null
}

Unknown or withheld values (17)

$.chaining not-stated
$.egress.metered egress-unknown
$.egress.terms not-stated
$.evidence[0].proved process
$.extras.keep_rate_basis process
$.extras.price_note process
$.identity_cost.detail defaulted
$.identity_cost.kind identity-defaulted
$.interface interface-unknown
$.interface_detail withheld
$.limits_note not-stated
$.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

Replicate (r8.im + Files API) Storage

fails: quotainstitutionno cardRper-object-capdisagrees

replicate.com · paas · storage/replicate · JSON · compare

Sibling records disagree:

identity_kind: storage/replicate says institution, storage/replicate-2 says email

Free tier	documented behaviour is expiry, not quota. r8.im is a private-ish OCI registry for cog models. unit mixed; metered on unstated
Price	none stated
Per-object cap	per-object-cap 100 MB (100000000 bytes)
Retention	not stated
Egress	Per-prediction output files are served from Replicate's CDN, but only for the retention window.
Identity	institution documented behaviour is expiry, not quota. r8.im is a private-is
Limits	per-object-cap per-object-cap 100 MB source per-object-cap 100 MB via the client library ('This will work for files up to 100MB'); data-URI path recommended only <1 MB. Retention is the killer: 'For predictions created through the API, all input parameters, output values, output files, and logs are automatically removed after an hour, by default.' Files API objects carry an explicit `expires_at`.
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
Tooling	`cog push r8.im/<owner>/<model>`, replicate-python Files API (files.create/list/get/delete).
Composes with	No viable chain: 100 MB per object x 1-hour retention. This is the textbook 'big per-file cap lives where retention is shortest' inverse — here small cap AND short retention.
Reliability	r8.im has no anonymous pull path I could find: GET /v2/ -> 401, /v2/token -> {'error':'Endpoint not found'}, and /v2/<repo>/manifests/latest 307s to a replicate.com HTML page that 404s.
Notes	documented behaviour is expiry, not quota. r8.im is a private-ish OCI registry for cog models.
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 100.0 MB

Evidence (3)

date	source	what it proved
2026-10-21	replicate.com/docs/topics/predictions/data-retention	'all input parameters, output values, output files, and logs are automatically removed after an hour, by default'
2026-10-21	replicate.com/docs/topics/predictions/input-files	'This will work for files up to 100MB'
2026-10-21	withheld	probe: GET r8.im/v2/ -> 401; GET r8.im/v2/token -> 404 {'error':'Endpoint not found'}; GET r8.im/replicate/cog-hello-world/manifests/latest -L -> 307 -> 404 HTML

More fields (1)

{
 "dead_end": {
  "failure_class": "quota",
  "why": "documented behaviour is expiry, not quota. r8.im is a private-ish OCI registry for cog models.",
  "retry_when": null,
  "first_recorded": "2026-10-21"
 }
}

Unknown or withheld values (6)

$.evidence[2].url private-url
$.extras.dead_end.retry_when process
$.free_tier.notes process
$.per_object_cap.notes not-extracted
$.retention.notes not-extracted
$.retention.stated not-extracted

replicate (model hosting) Storage

fails: quotaemail?card?Rdisagrees

replicate.com/docs/topics/predictions/output-files · rejected-provider · storage/replicate-2 · JSON · compare

Sibling records disagree:

identity_kind: storage/replicate says institution, storage/replicate-2 says email

Free tier	Output files are ephemeral by design: 'output files are automatically deleted after an hour' for API-created predictions, and replicate.delivery URLs 'expire after one hour'. Custom models a 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	Output files are ephemeral by design: 'output files are automatically deleted after an hour' for API-created predictions, and replicate.delivery URLs 'expire after one hour'. Custom models a
Notes	Output files are ephemeral by design: 'output files are automatically deleted after an hour' for API-created predictions, and replicate.delivery URLs 'expire after one hour'. Custom models a
Evidence rung	reported
Checked	2026-10-21 (inherited); stale on 2026-11-04

Evidence (2)

date	source	what it proved
2026-10-21	replicate.com/docs/topics/predictions/output-files	Output files are ephemeral by design: 'output files are automatically deleted after an hour' for API-created predictions, and replicate.delivery URLs 'expire after one hour'. Custom models a
2026-10-21	replicate.com/pricing	cited as the basis of the rejection: Output files are ephemeral by design: 'output files are automatically deleted after an hour' for API-created predictions, and replicate.delivery URLs 'expire after one hour'. Custom models a

More fields (1)

{
 "dead_end": {
  "failure_class": "retention",
  "why": "Output files are ephemeral by design: 'output files are automatically deleted after an hour' for API-created predictions, and replicate.delivery URLs 'expire after one hour'. Custom models a",
  "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