Featurize
featurize.cn · Compute · 1 record · JSON
Featurize Compute
fails: identitycountry-phone?card?phoneDwrite-gated
| Free tier | Chinese GPU cloud and hosted notebook with published per-hour rates: RTX 4090 quoted at CNY 1.87 per hour on 16 AMD EPYC 9354 cores, 60 GB RAM and 750 GB of free disk; RTX 3090 quoted at CNY 1.62 per hour on 6 Xeon Gold 6142 cores, 60 GB RAM and 450 GB disk; PRO 6000 at CNY 6.19 per hour. At USD 1 = CNY 6.6743 that is USD 0.280, USD 0.243 and USD 0.927 per GPU-hour. PyTorch and TensorFlow images are preinstalled and billing is on demand or by monthly subscription. unit hours; metered on wall-clock; none |
|---|---|
| Price | none stated |
| Per-object cap | not applicable |
| Retention | not applicable |
| Egress | not stated |
| Identity | country-phone (placeholder) unverified, inferred: https://featurize.cn/register answers 404, so no registration form was read. A mainland number is inferred from the price currency and from every sibling service in this class using SMS registration; it is not quoted. |
| Limits |
The account is a Chinese mobile number plus an SMS code and the site publishes no alternate registration route, so the same identity wall as AutoDL applies. |
| Binding | write-gated The account is a Chinese mobile number plus an SMS code and the site publishes no alternate registration route, so the same identity wall as AutoDL applies. |
| Interface | rest inferred from the tooling the provider documents; not probed |
| Tooling | not stated |
| Composes with | The 750 GB of free instance disk alongside a 4090 is larger than any Western broker's included disk at this rate, which matters for a dataset-heavy fine-tune. |
| Reliability | no reliability event recorded |
| Notes | New provider: no record in dossier/compute/. The phone gate is inferred from the site being a mainland consumer service and from AutoDL's sibling form; the Featurize registration page was not read, so that is the weaker claim in this record and it is flagged in binding_note rather than presented as quoted. If an email route exists, this would be one of the cheapest 24 GB GPU hours in the corpus. |
| Evidence rung | documented |
| Checked | 2026-10-22 (live); stale on 2026-11-21 |
Evidence (2)
| date | source | what it proved |
|---|---|---|
| 2026-10-22 | featurize.cn/ | translated from the Chinese page: the RTX 4090 row reads on-demand as low as CNY 1.87 per hour, on 16 AMD EPYC 9354 cores, 60 GB of high-speed memory and 750 GB of free disk; the RTX 3090 row reads CNY 1.62 per hour; the PRO 6000 row reads CNY 6.19 per hour |
| 2026-10-22 | api.frankfurter.app/latest?from=USD&to=CNY | USD 1 = CNY 6.6743 on 2026-10-22 |
More fields (1)
{
"wave2": {
"class": "regional-gpu-retailer",
"free_tier_kind": "none",
"shape": {
"vcpu": 16,
"ram_gib": 60,
"gpu_class": "RTX 4090",
"gpu_count": 1,
"vram_gib": 24,
"disk_gib": 750
}
}
}Unknown or withheld values (9)
$.egress.meteredegress-unknown$.egress.termsnot-stated$.identity_cost.kindidentity-defaulted$.per_object_cap.notesprocess$.per_object_cap.statednot-applicable$.reliability_notenone-observed$.retention.notesprocess$.retention.statednot-applicable$.toolingnot-stated