Lightning AI

lightning.ai · Compute, Storage · 3 records · JSON

Lightning AI Studios Storage★ qualifiescard?cardRretention-clamped
Lightning AI (Lightning Studios persistent storage) Storagefails: quotaemailno cardR
Lightning AI Computedeademailno cardDwrite-gated

Lightning AI Studios Storage

★ qualifiescard?cardRretention-clamped

lightning.ai · iaas · storage/lightning-ai-studios · JSON · compare

Free tier	Free plan: '1 free active Studio (4 hour restarts)', 'Persistent storage (200 GB limit)'. Studio docs say the default Studio disk is 400 GB; pricing page caps the free tier at 200 GB. unit GB; metered on stored-bytes
Price	none stated
Per-object cap	not stated
Retention	Free Studio restarts every 4 hours (compute restarts, disk persists); auto-sleep after 10 min idle; free GPU hours come from 'up to 30 credits'; account data deleted after 6 months without login.
Egress	No egress meter or per-GB transfer price is published on the pricing page — free/public app serving on Studios is included in the plan.
Identity	card (placeholder) inferred from the category 'iaas'; the source record states no identity cost, so this is a class assumption, not evidence
Limits	retention-clamped deleted after 6 months source Free Studio restarts every 4 hours (compute restarts, disk persists); auto-sleep after 10 min idle; free GPU hours come from 'up to 30 credits'; account data deleted after 6 months without login.
Binding	retention-clamped a clock, not a quota: bytes leave whether or not the client is finished with them
Interface	rest, sftp, ssh, rclone, cli rest
Tooling	SSH/SFTP into a Studio (docs: 'Connect any local IDE or SSH'), plus a documented public-port feature; `lightning` CLI and SDK. No rclone backend.
Composes with	1 Studio per free account with 200 GB persistent; N free accounts. Studios can be mounted as read-only 'datasets' shared across a teamspace, so one account can fan out many consumers.
Reliability	Disk genuinely persists across sleep/switch ('Studios persist their environments when sleeping and switching machines'); the 4-hour restart interrupts serving but does not lose files.
Notes	Free plan: '1 free active Studio (4 hour restarts)', 'Persistent storage (200 GB limit)'. Studio docs say the default Studio disk is 400 GB; pricing page caps the free tier at 200 GB.
Evidence rung	reported One page read moves it to documented.
Checked	2026-10-21 (inherited); stale on 2026-11-04

Evidence (4)

date	source	what it proved
2026-10-21	lightning.ai/pricing	Free plan list: '1 free active Studio (4 hour restarts)', 'Persistent storage (200 GB limit)'
2026-10-21	lightning.ai/docs/platform/build/ai-studio/environment-persi	'Everything in the Studio home (~, /teamspace/studios/this_studio) is persisted'
2026-10-21	lightning.ai/docs/platform/build/ai-studio/deploy-on-public-	'Studios support any kind of server as long as it exposes a port'
2026-10-21	lightning.ai/docs/platform/security/privacy/data-retention-p	6-month inactivity -> account scheduled for deletion, 30-day notice

Unknown or withheld values (5)

$.free_tier.notes process
$.identity_cost.kind identity-defaulted
$.per_object_cap.notes not-extracted
$.per_object_cap.stated not-extracted
$.retention.notes not-extracted

Lightning AI (Lightning Studios persistent storage) Storage

fails: quotaemailno cardR

lightning.ai · iaas · storage/lightning-ai · JSON · compare

Free tier	No GB figure published. The free tier is framed entirely in GPU hours ('Start on the free tier with 80* free GPU hours'), one Studio free 24/7 with a forced restart every 4 hours, and 'Studios include persistent storage, allowing file and data downloads to remain intact even if the instance stops, without incurring costs'. So storage is included-and-unmetered in prose, but with no stated ceiling. unit GB; metered on stored-bytes
Price	none stated
Per-object cap	not stated
Retention	not stated
Egress	Not metered — 'without incurring costs' is the only storage statement, and there is no egress line.
Identity	email email (GitHub OAuth).
Binding	none no constraint of a named kind was recorded; the verdict rests on the quota sentence rather than on a named limit
Interface	rest, cli rest
Tooling	`lightning` CLI, `lightning studio` mount/sync, and a Python SDK; Studios are Docker containers with a persistent /teamspace. Real client, real product.
Composes with	One free Studio runs 24/7 per account, so N accounts x 1 persistent volume. The ceiling per volume is undisclosed, which makes planning impossible.
Reliability	The 4-hour forced restart is a hard operational constraint, and the absence of any written storage ceiling means the number can move without notice — the brief's 'client-side limits are worthless as evidence' rule applies in reverse here: there is no server-side limit published either.
Notes	No GB figure published. The free tier is framed entirely in GPU hours ('Start on the free tier with 80* free GPU hours'), one Studio free 24/7 with a forced restart every 4 hours, and 'Studios include persistent storage, allowing file and data downloads to remain intact even if the instance stops, without incurring costs'. So storage is included-and-unmetered in prose, but with no stated ceiling.
Evidence rung	reported One page read moves it to documented.
Checked	2026-10-21 (inherited); stale on 2026-11-04

Evidence (2)

date	source	what it proved
2026-10-21	lightning.ai/pricing	read live: 'Get up to 80 free GPU hours to start'; 'Run 1 Studio free 24/7 ... Free Studios run 24/7 but require restart every 4 hours'
2026-10-21	lightning.ai/docs/platform/build/ai-studio	read live (AI Studio docs): 'Studios include persistent storage, allowing file and data downloads to remain intact even if the instance stops, without incurring costs'

More fields (1)

{
 "dead_end": {
  "failure_class": "egress",
  "why": "No GB figure published. The free tier is framed entirely in GPU hours ('Start on the free tier with 80* free GPU hours'), one Studio free 24/7 with a forced restart every 4 hours, and 'Studios include persistent storage, allowing file and data downloads to remain intact even if the instance stops, without incurring costs'. So storage is included-and-unmetered in prose, but with no stated ceiling.",
  "retry_when": null,
  "first_recorded": "2026-10-21"
 }
}

Unknown or withheld values (6)

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

Lightning AI Compute

deademailno cardDwrite-gated

lightning.ai/pricing · dev environment · compute/lightning-ai · JSON · compare

Free tier	one Studio free 24/7, no credit card, no commitment; free Studios require a restart every 4 hours; 80 free GPU hours to start unit hours; metered on wall-clock; the 80 GPU hours are a starting grant against a table from T4 to H200
Price	none stated
Per-object cap	not applicable
Retention	a Studio keeps its files between sessions; the free tier forces a restart every 4 hours
Egress	not priced on the page read
Identity	email 'No credit card' for the free Studio, quoted
Limits	write-gated all user access revoked on 2026-10-16 01:00 UTC; no Studio can be created while the incident is open source the 4-hour restart is the constraint: a process that must survive it needs a supervisor that resumes after the restart
Binding	write-gated the platform is closed to all users while an incident is open: no Studio can be created, so the free allowance cannot be reached at all
Interface	rest, cli web IDE, SSH into a Studio, and a CLI
Tooling	lightning CLI, SSH, the Studio IDE; persistent filesystem per Studio
Composes with	the free CPU Studio is the only free 24/7 machine with a persistent disk and no card
Reliability	access revoked for every user on 2026-10-16 at 01:00 UTC after credential compromise; the incident page states the platform stays inaccessible to all users and status.lightning.ai still reported the service down on 2026-10-22
Notes	The offer is real on the pricing page and the platform is not: the security-incident page still states the Platform remains inaccessible, and the status banner still reads 'Service is down'. A lane on 2026-10-24 read only the pricing page and reversed this verdict to qualifies; that reading is refuted by the two operational pages read the same day. Keep dead. Reopen when status.lightning.ai clears the banner and the incident page announces restoration.
Evidence rung	documented
Checked	2026-10-25 (live); stale on 2026-11-24
Parsed numbers	`$.extras.m10h_usd` price USD0.0 / month `$.extras.m24h_usd` price USD0.0 / month

Evidence (11)

date	source	what it proved
2026-10-22	lightning.ai/pricing	'Run 1 Studio free 24/7. No credit card. No commitments.'
2026-10-22	lightning.ai/pricing	'Free Studios run 24/7 but require restart every 4 hours. No restrictions in Pro or higher.'
2026-10-22	lightning.ai/pricing	80 free GPU hours; T4 USD 0.55/h, L4 USD 0.79, L40S USD 3.54, A100 40GB USD 2.19, H100 USD 5.45, H200 USD 4.50
2026-10-22	lightning.ai/security-incident	'We revoked all user access to the Lightning Platform on October 16 at 1:00 AM UTC' and 'the Platform will remain inaccessible to all users', quoted; affected services named as Studio, Deployments, Pipelines and Drive
2026-10-22	status.lightning.ai	banner reads 'Service is down to address a security incident', last update 2026-10-19 12:49 UTC, incident still open when read
2026-10-22	lightning.ai/pricing	the pricing page still advertises 'Run 1 Studio free 24/7', so the marketing page contradicts the incident page and the incident page is the newer and more specific claim
2026-10-23	lightning.ai/pricing	'You get 5 free Lightning credits upon registration. Add a card for 25 more. If you don't use them, they expire in 12 months. Credits can be used on compute, storage and tokens.', quoted on 2026-10-23; this conflicts with the recorded revocation of user access on 2026-10-16 and does not overturn it
2026-10-24	lightning.ai/security-incident	quoted: 'This is an active investigation and during this time the Platform will remain inaccessible to all users'
2026-10-24	status.lightning.ai	banner reads 'Service is down to address a security incident'; the newest update is dated 2026-10-19 12:49 UTC
2026-10-24	lightning.ai/pricing	the marketing page still advertises 'Run 1 Studio free 24/7. No credit card. No commitments.' while the platform is down
2026-10-24	lightning.ai/pricing	the quoted rate table, all eight rows: T4 USD 0.55 with 75 free hours to start, L4 USD 0.79 with 31, L40S USD 3.54 with 5, RTXP 6000 USD 4.64 with 3, A100 40 GB USD 2.19 with 10, A100 80 GB USD 1.85 with none, H100 USD 5.45 with none and H200 USD 4.50 with none; footnotes read 'Billed by the second' and 'Free Studios run 24/7 but require restart every 4 hours'

More fields (14)

{
 "shape": {
  "vcpu": null,
  "ram_gib": null,
  "disk_gib": null,
  "gpu_class": null,
  "gpu_count": null
 },
 "billing_model": "wall-clock",
 "granularity": {
  "min_billed_s": 60,
  "rounding_s": 60
 },
 "session_cap": {
  "wall_clock_min": 240,
  "idle_reap_s": null,
  "restart_required": true,
  "note": "a free Studio must restart every 4 hours; Pro removes the restriction"
 },
 "keep_rate": {
  "value": 0.0,
  "basis": "a stopped Studio bills storage only",
  "source": "https://lightning.ai/pricing"
 },
 "persistent_storage": {
  "included_gib": null,
  "price_gib_month": "not stated on the pricing page",
  "durability": "Studio filesystem persists",
  "survives_stop": true,
  "survives_destroy": null,
  "max_retention": "while the Studio exists"
 },
 "network_out": {
  "allow_outbound": true,
  "allow_inbound_listen": null,
  "ports": null,
  "static_ip": null,
  "ipv6": null,
  "per_session_bytes": null
 },
 "free_tier_kind": "allowance",
 "m10h_usd": 0.0,
 "m24h_usd": 0.0,
 "price_arithmetic": "the CPU Studio is free at any duty cycle; the 4-hour restart is the cost",
 "census_rows": [
  "lightning-ai"
 ],
 "burst_x": 1.0,
 "wave2": {
  "verdict_change": "qualifies-with-caveat -> dead on 2026-10-22"
 }
}

Unknown or withheld values (2)

$.per_object_cap.notes process
$.per_object_cap.stated not-applicable

Corrections

2026-10-22 Lightning AI's free Studio is dead, and the pricing page still sells it