Synteq

gpurentalprices.com · Compute · 4 records · JSON

Hot Aisle Computepaidemailno cardR
Prime Intellect Marketplace Computepaidemailno cardRvolume-quota
SaladCloud community floor Computepaidemailno cardRretention-clamped
Synteq ComputepaidcardcardRwrite-gated

Hot Aisle Compute

paidemailno cardR

gpurentalprices.com/providers/hotaisle · GPU cloud · compute/hot-aisle · JSON · compare

Free tier	AMD-only GPU host, one model: MI300X at USD 2.99 per GPU-hour on demand, checked 2026-10-22 and unchanged since 2026-10-15. Third-party billing facts, each attributed to the vendor's own pages: billing per minute, minimum rental of one GPU for one minute, and prepaid credits bought by card through Stripe or in USDT and USDC. unit hours; metered on wall-clock; none
Price	none stated
Per-object cap	not applicable
Retention	not applicable
Egress	not stated
Identity	email Prepaid credits rather than a subscription, and the tracker records payment by card or by USDT and USDC, so a stablecoin balance is an alternative to a card.
Binding	none Nothing published binds above a single GPU for a single minute; the whole cost is prepaid and the smallest unit is one minute.
Interface	unknown inferred from the tooling the provider documents; not probed
Tooling	not stated
Composes with	The one-minute minimum and per-minute billing make this the only 192 GB card sold in units a benchmark can afford, which is what a trial run needs before a multi-hour commitment.
Reliability	no reliability event recorded
Notes	New provider: no record in dossier/compute/. Storage price, data-out price and the rule for stopped instances are stated by the tracker as unpublished by the vendor, which matters because a 192 GB card implies a large dataset beside it.
Evidence rung	reported
Checked	2026-10-22 (live); stale on 2026-11-05

Evidence (3)

date	source	what it proved
2026-10-22	gpurentalprices.com/providers/hotaisle	quoted: 'Hot Aisle rents 1 GPU model today, from $2.99 an hour (MI300X, on-demand)'; source of the rate given as hotaisle.xyz/pricing
2026-10-22	gpurentalprices.com/providers/hotaisle	quoted billing terms: 'Billing is per minute', 'Minimum rental time: one GPU for one minute', 'prepaid credits by card (Stripe) or USDT and USDC', each checked on 2026-10-17
2026-10-22	gpurentalprices.com/mi300x	MI300X market median across 5 providers is USD 2.99/hr, so Hot Aisle matches the median rather than undercutting it

More fields (1)

{
 "wave2": {
  "class": "gpu-broker-rate-floor",
  "free_tier_kind": "none",
  "shape": {
   "vcpu": null,
   "ram_gib": null,
   "gpu_class": "MI300X",
   "gpu_count": 1,
   "vram_gib": 192
  }
 }
}

Unknown or withheld values (9)

$.egress.metered egress-unknown
$.egress.terms not-stated
$.interface interface-unknown
$.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

Prime Intellect Marketplace Compute

paidemailno cardRvolume-quota

gpurentalprices.com/providers/primeintellect · GPU cloud · compute/prime-intellect-marketplace · JSON · compare

Free tier	Marketplace lane of Prime Intellect, priced per GPU-hour: H200 141GB at USD 3.14, B200 180GB at USD 3.49, B300 288GB at USD 4.99, and spot or interruptible H100 at USD 0.94, which is the cheapest H100 hour recorded anywhere. Prices checked 2026-10-22. unit hours; metered on wall-clock; none
Price	none stated
Per-object cap	not applicable
Retention	not applicable
Egress	not stated
Identity	email Unchanged from the existing record: the platform uses prepaid credits, and the grant programme that once existed is gone.
Limits	volume-quota Prepaid credits are the ceiling: the existing record quotes the documentation, credits are deducted every minute while a Pod is active and Pods stop when credits run out, so the balance is what limits a run. source Prepaid credits are the ceiling: the existing record quotes the documentation, credits are deducted every minute while a Pod is active and Pods stop when credits run out, so the balance is what limits a run.
Binding	volume-quota Prepaid credits are the ceiling: the existing record quotes the documentation, credits are deducted every minute while a Pod is active and Pods stop when credits run out, so the balance is what limits a run.
Interface	unknown inferred from the tooling the provider documents; not probed
Tooling	not stated
Composes with	USD 0.94 for an interruptible H100 hour is half the cheapest on-demand H100 in the same market, so a checkpointing run that tolerates preemption gets two H100 hours for the price of one anywhere else.
Reliability	no reliability event recorded
Notes	Extends the existing prime-intellect.json record, which covers the sandbox pricing only, and prime-intellect-2.json, which records the grant programme as dead. The marketplace rates are third-party reported and the vendor pricing page answered 404 on 2026-10-22, so the rung is reported. A spot rate is not a supply guarantee: the tracker labels it spot, and the existing record's note that Pods stop when credits run out is the practical limit.
Evidence rung	reported
Checked	2026-10-22 (live); stale on 2026-11-05

Evidence (2)

date	source	what it proved
2026-10-22	gpurentalprices.com/providers/primeintellect	the tracker names Prime Intellect as the spot floor for H100 at USD 0.94 per GPU-hour and as the on-demand floor for H200 at USD 3.14, B200 at USD 3.49 and B300 at USD 4.99
2026-10-22	gpurentalprices.com/	quoted: 'Spot from $0.94 (Prime Intellect)' and 'Cheapest on-demand H100 today: $1.80/hr at HyperAI' with the H100 median at USD 3.29

More fields (1)

{
 "wave2": {
  "class": "gpu-marketplace",
  "free_tier_kind": "none",
  "shape": {
   "vcpu": null,
   "ram_gib": null,
   "gpu_class": "H100",
   "gpu_count": 1,
   "vram_gib": 80
  }
 }
}

Unknown or withheld values (9)

$.egress.metered egress-unknown
$.egress.terms not-stated
$.interface interface-unknown
$.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

SaladCloud community floor Compute

paidemailno cardRretention-clamped

gpurentalprices.com/providers/salad · GPU cloud · compute/saladcloud-community-floor · JSON · compare

Free tier	The cheapest whole-card hour recorded: an RTX 3090 24GB at USD 0.09 per GPU-hour on the community tier, an RTX 4090 24GB at USD 0.16, an RTX 5090 32GB at USD 0.25 and an L40S 48GB at USD 0.79, checked 2026-10-22. These are interruptible community nodes, which the vendor's own pricing page prices by priority tier, with medium and high priority priced above the lowest tier. unit hours; metered on wall-clock; none
Price	none stated
Per-object cap	not applicable
Retention	Community nodes are the lowest priority tier and are preempted by higher-priority workloads, so a job that must run to completion cannot rely on holding the card.
Egress	not stated
Identity	email Unchanged from the existing record: an account, with no free trial and no minimum commitment quoted from the vendor FAQ.
Limits	retention-clamped Community nodes are the lowest priority tier and are preempted by higher-priority workloads, so a job that must run to completion cannot rely on holding the card. source Community nodes are the lowest priority tier and are preempted by higher-priority workloads, so a job that must run to completion cannot rely on holding the card.
Binding	retention-clamped Community nodes are the lowest priority tier and are preempted by higher-priority workloads, so a job that must run to completion cannot rely on holding the card.
Interface	unknown inferred from the tooling the provider documents; not probed
Tooling	not stated
Composes with	At USD 0.09 a 3090-hour, USD 10 buys 111 hours on a 24 GB card, which is the cheapest way to run a long checkpointed fine-tune the preemption risk is what the discount pays for.
Reliability	no reliability event recorded
Notes	Corrects the existing salad.json record, whose offer field leads with 'GPUs from USD 0.015 per hour'. That figure is the vendor's marketing floor and does not correspond to a whole card on the community tier: the cheapest named 24 GB card is USD 0.09 for a 3090. The vendor page renders its per-class table client-side, so the numbers came from the tracker rather than from the vendor's own HTML, which is why the rung is reported.
Evidence rung	reported
Checked	2026-10-22 (live); stale on 2026-11-05

Evidence (4)

date	source	what it proved
2026-10-22	gpurentalprices.com/rtx-3090	quoted: RTX 3090 spot or marketplace from USD 0.09 'community, SaladCloud' across 3 providers
2026-10-22	gpurentalprices.com/rtx-4090	quoted: RTX 4090 spot or marketplace from USD 0.16 'community, SaladCloud'
2026-10-22	gpurentalprices.com/rtx-5090	quoted: RTX 5090 spot or marketplace from USD 0.25 'community, SaladCloud', down 12.8 percent in 7 days
2026-10-22	salad.com/pricing	quoted page text: four priority tiers where 'Higher priority can take over lower-priority workloads', and 'Lowest is the lowest cost and the most interruptible'

More fields (1)

{
 "wave2": {
  "class": "gpu-marketplace",
  "free_tier_kind": "none",
  "shape": {
   "vcpu": null,
   "ram_gib": null,
   "gpu_class": "RTX 3090",
   "gpu_count": 1,
   "vram_gib": 24
  }
 }
}

Unknown or withheld values (8)

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

Synteq Compute

paidcardcardRwrite-gated

gpurentalprices.com/providers/synteq · GPU cloud · compute/synteq · JSON · compare

Free tier	Monthly-prepaid GPU host with four listed models and nothing in stock as of 2026-10-22. Cheapest listed configuration is an RTX PRO 4000 Blackwell at USD 185 per month, which the tracker divides to USD 0.25 per GPU-hour over 730 hours. Terms quoted from the vendor's own documentation: block storage USD 4 per TB per month, bandwidth included on every plan, billing monthly and prepaid, minimum rental one month, and cancelling gives no partial refund. unit hours; metered on wall-clock; none
Price	none stated
Per-object cap	not applicable
Retention	not applicable
Egress	quoted: 'None of these listed configurations is reported in stock'; and the terms 'Billing is monthly, prepaid', 'Minimum rental time: 1 month; cancelling gives no partial refund', 'Data out: bandwidth included on every plan'
Identity	card The first month is paid up front and the tracker records no free tier.
Limits	write-gated A one-month prepaid commitment with no partial refund is the binding term: a cancelled month is a lost month, and none of the four configurations was in stock on the day checked. source A one-month prepaid commitment with no partial refund is the binding term: a cancelled month is a lost month, and none of the four configurations was in stock on the day checked.
Binding	write-gated A one-month prepaid commitment with no partial refund is the binding term: a cancelled month is a lost month, and none of the four configurations was in stock on the day checked.
Interface	unknown inferred from the tooling the provider documents; not probed
Tooling	not stated
Composes with	A USD 185 month with bandwidth included is a fixed-cost way to hold one card continuously, but LeaderGPU's USD 140 month for a 3090 is cheaper and this is a commitment, so it only composes with a workload that runs for most of 730 hours.
Reliability	no reliability event recorded
Notes	New provider: no record in dossier/compute/. Supply is the finding and it is explicit: zero in stock. The USD 0.25 per hour is the tracker's division of a monthly price, not a published hourly rate, and a month used for 100 hours costs USD 1.85 per hour on the same invoice. Recorded because the monthly term and the no-refund policy are exactly the shape this corpus lacks, not because the capacity is currently reachable.
Evidence rung	reported
Checked	2026-10-22 (live); stale on 2026-11-05

Evidence (3)

date	source	what it proved
2026-10-22	gpurentalprices.com/providers/synteq	quoted: 'Synteq rents 4 GPU models today, from $0.25 an hour (RTX PRO 4000 Blackwell, reserved / committed, 1 month, $185/month divided by 730 hours; monthly billing, out of stock)'
2026-10-22	gpurentalprices.com/providers/synteq	quoted: 'None of these listed configurations is reported in stock'; and the terms 'Billing is monthly, prepaid', 'Minimum rental time: 1 month; cancelling gives no partial refund', 'Data out: bandwidth included on every plan'
2026-10-22	synteq.com/api/catalog/vps	the tracker names this endpoint as the source of the catalogue it reads, which is the machine-readable way to re-check availability later

More fields (1)

{
 "wave2": {
  "class": "committed-gpu-host",
  "free_tier_kind": "none",
  "shape": {
   "vcpu": null,
   "ram_gib": null,
   "gpu_class": "RTX PRO 4000 Blackwell",
   "gpu_count": 1,
   "vram_gib": null
  }
 }
}

Unknown or withheld values (7)

$.interface interface-unknown
$.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