Qualcomm AI

qualcomm.com · Compute · 2 records · JSON

Qualcomm AI Hub Computecaveatemailno cardphoneD
Qualcomm Device Cloud Computecaveatemailno cardRvolume-quota

Qualcomm AI Hub Compute

caveatemailno cardphoneD

workbench.aihub.qualcomm.com/docs/hub/faq.html · cloud VM · compute/qualcomm-ai-hub · JSON · compare

Free tier	Free compile, profiling and inference on real Qualcomm devices, in the cloud: 'Qualcomm AI Hub Workbench, our platform, is currently completely free to use. We encourage you to submit your models, compile them, and submit inference jobs to both iterate on performance and to check the accuracy and download the target asset'. The workflow is described as 'submit an inference job, running your compiled model on a real device hosted in the cloud', with per-layer timing and runtime logs, and a second job uploads input data and downloads outputs. unit GiB; metered on wall-clock; free-forever
Price	none stated
Per-object cap	not applicable
Retention	not applicable
Egress	not stated
Identity	email a Qualcomm MyAccount plus an API token taken from Account then Settings, with no payment step described in the getting-started docs, so no card and no hold on that path. Whether the MyAccount signup asks for a phone number is not stated in the evidence, and debit and prepaid acceptance are not stated because nothing is charged
Binding	none Qualcomm publishes no request rate, job count, queue depth or model size ceiling beyond a note that models above 2 GB may fail to compile, so the constraint that fails first cannot be named from the documentation.
Interface	ssh, rest inferred from the tooling the provider documents; not probed
Tooling	not stated
Composes with	It chains with the Qualcomm Device Cloud record already in this corpus: AI Hub gives the compile and profile loop on hosted silicon, Device Cloud gives an interactive session with SSH port forwarding on a Snapdragon X Elite machine. Together they cover model work and application work without owning a phone.
Reliability	no reliability event recorded
Notes	The one gap is the ceiling: 'completely free' with no published quota means the honest reading is free until Qualcomm changes it, and a heavy user could discover a limit the docs never state. The device list page warns that some entries are proxy devices that 'do not host the actual device', so per-model metrics may be modelled rather than measured; the inference-job path explicitly runs on real hardware, and that is the path this finding rests on. No probe was run against the API, so the verdict is qualifies-with-caveat rather than qualifies: the free claim is a vendor FAQ sentence, and the row would need an account, a token and one inference job before it could be called measured.
Evidence rung	documented
Checked	2026-10-22 (live); stale on 2026-11-21

Evidence (4)

date	source	what it proved
2026-10-22	workbench.aihub.qualcomm.com/docs/hub/faq.html	'Qualcomm AI Hub Workbench, our platform, is currently completely free to use', quoted, in the fees question
2026-10-22	workbench.aihub.qualcomm.com/docs/hub/faq.html	'submit an inference job, running your compiled model on a real device hosted in the cloud and analyze the performance', quoted
2026-10-22	workbench.aihub.qualcomm.com/docs/hub/getting_started.html	access is a Qualcomm MyAccount plus an API token from Account then Settings, with no payment step described
2026-10-22	workbench.aihub.qualcomm.com/docs/hub/faq.html	compilation limits: 'The model is large (i.e. > 2GB)' is one named cause of a failed compile

More fields (1)

{
 "wave2": {
  "class": "hardware-vendor-programme",
  "free_tier_kind": "free-forever",
  "shape": {
   "vcpu": null,
   "ram_gib": null,
   "gpu_class": "Qualcomm Hexagon NPU / Adreno GPU on hosted Snapdragon devices",
   "gpu_count": 1
  }
 }
}

Unknown or withheld values (8)

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

Qualcomm Device Cloud Compute

caveatemailno cardRvolume-quota

qdc.qualcomm.com/support/faq · dev environment · compute/qualcomm-device-cloud · JSON · compare

Free tier	Free to use: create a Qualcomm ID, then request free minutes for a device and run interactive sessions or automated jobs on real devices, with SSH port forwarding into the session. The FAQ states new users get free pre-provisioned minutes on Snapdragon X Elite on first login for a limited time. A third-party guide states each device can be accessed free for up to 5,000 minutes. unit minutes; metered on wall-clock; recurring-quota
Price	none stated
Per-object cap	not applicable
Retention	not applicable
Egress	not stated
Identity	email not published
Limits	volume-quota minutes are a pooled allocation requested per device, and an organisation's minutes are shared across its users, so the 5,000-minute figure is per device rather than per account and does not grow with the number of machines; further minutes require a reviewed request. source minutes are a pooled allocation requested per device, and an organisation's minutes are shared across its users, so the 5,000-minute figure is per device rather than per account and does not grow with the number of machines; further minutes require a reviewed request.
Binding	volume-quota minutes are a pooled allocation requested per device, and an organisation's minutes are shared across its users, so the 5,000-minute figure is per device rather than per account and does not grow with the number of machines; further minutes require a reviewed request.
Interface	ssh inferred from the tooling the provider documents; not probed
Tooling	not stated
Composes with	a real Snapdragon X Elite on ARM64 with SSH is the only lane in the corpus for testing ARM64 build output and on-device NPU inference without buying hardware, and it pairs with a free CI runner that cross-compiles; the minutes pool is what breaks, and it is consumed by interactive time rather than by CPU work.
Reliability	no reliability event recorded
Notes	Absent from dossier/compute/ (checked against 2,395 provider names; no Qualcomm record exists). The '5,000 minutes' figure is third-party, so the record is rung `reported` while the free-to-use fact is documented on the vendor FAQ. Device minutes are shared at organisation level, which is worth knowing before relying on the number for a long job. Not probed: no Qualcomm ID was created, so the actual allowance on first login is unverified.
Evidence rung	reported
Checked	2026-10-22 (live); stale on 2026-11-05

Evidence (3)

date	source	what it proved
2026-10-22	qdc.qualcomm.com/support/faq	'Is Qualcomm Device Cloud free to use? Yes.'; usage requires a Qualcomm ID; users 'request free minutes for a device'; minutes start counting only once test execution begins; additional minutes are provided after a reviewed request, quoted
2026-10-22	qdc.qualcomm.com/support/faq	'We are currently promoting Snapdragon X Elite devices, and new users get free pre-provisioned minutes on Snapdragon X Elite upon first login for a limited time', quoted
2026-10-22	github.com/globaledgesoft/Provisioning-Snapdragon-X-Elite-on	'each device can be accessed for free for up to 5000 minutes', and a form is used to unlock the free 5,000 minutes

More fields (1)

{
 "wave2": {
  "class": "perpetual-free-allowance",
  "free_tier_kind": "recurring-quota",
  "shape": {
   "vcpu": null,
   "ram_gib": null,
   "gpu_class": "Snapdragon X Elite GPU",
   "gpu_count": 0
  }
 }
}

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