PPIO GPU
ppio.com · Compute · 2 records · JSON
- PPIO GPU Computepaidinstitutionno cardphoneD
- PPIO Agent Sandbox Computepaidemail?card?Dwrite-gated
PPIO GPU Compute
paidinstitutionno cardphoneD
| Free tier | RTX 4090 24GB in a 16 vCPU / 62 GiB container at CNY 2.31 per card-hour on demand or CNY 1.16 per card-hour preemptible; H20 96GB at CNY 6.80 or CNY 3.40; H100 SXM 80GB at CNY 23.76 or CNY 11.88; H200 141GB listed as new stock. Storage is free to 100 GB and the page states preemptible capacity saves about 50 percent, quoted. unit GiB; metered on wall-clock; none |
|---|---|
| Price | none stated |
| Per-object cap | not applicable |
| Retention | not applicable |
| Egress | not stated |
| Identity | institution The GPU page links to ppio.com/user/register, but the form fields were not read, so whether a mainland mobile number or a real-name check is required is not established. The provider is a mainland company with an international storefront. |
| Binding | none Nothing is metered except wall-clock GPU time. The spot lane is the first thing to fail, because a preemptible 4090 at CNY 1.16 is reclaimable and the on-demand price is exactly double. |
| Interface | unknown inferred from the tooling the provider documents; not probed |
| Tooling | not stated |
| Composes with | A preemptible 4090 at CNY 1.16 an hour with 100 GB of free storage is the cheapest way to run a fine-tune that checkpoints often; the reclaim is the price of it. |
| Reliability | no reliability event recorded |
| Notes | dossier/compute/ holds ppio-sandbox.json, which is the E2B-compatible agent sandbox and carries no GPU price, so this is new capacity at a provider the corpus knows only for sandboxes. The aggregator's 0 percent preemption figure is a measurement of the aggregator on 2026-10-22, not a guarantee from the vendor. |
| Evidence rung | documented |
| Checked | 2026-10-22 (live); stale on 2026-11-21 |
Evidence (3)
| date | source | what it proved |
|---|---|---|
| 2026-10-22 | ppio.com/ai-computing/gpu-instance | price cards for RTX 4090 24GB at ¥2.31/卡时 on demand and ¥1.16/卡时 preemptible, H20 96GB at ¥6.80 and ¥3.40, H100 SXM 80GB at ¥23.76 and ¥11.88, quoted |
| 2026-10-22 | ppio.com/ai-computing/gpu-instance | page states 免费存储 100GB (100 GB of free storage) and 成本节省50% for preemptible capacity, quoted |
| 2026-10-22 | www.suanli.ai/gpus/RTX-4090 | an independent price aggregator lists PPIO 4090 at USD 0.346 on demand and USD 0.174 spot in CN, with an observed preemption rate of 0 percent, consistent with the vendor figures at CNY 6.67 per USD |
More fields (1)
{
"wave2": {
"class": "regional-floor-price",
"free_tier_kind": "none",
"shape": {
"vcpu": 16,
"ram_gib": 62,
"gpu_class": "RTX 4090",
"gpu_count": 1,
"vram_gib": 24,
"disk_gib": null
}
}
}Unknown or withheld values (9)
$.egress.meteredegress-unknown$.egress.termsnot-stated$.interfaceinterface-unknown$.per_object_cap.notesprocess$.per_object_cap.statednot-applicable$.reliability_notenone-observed$.retention.notesprocess$.retention.statednot-applicable$.toolingnot-stated
PPIO Agent Sandbox Compute
paidemail?card?Dwrite-gated
| Free tier | no free tier: vCPU and RAM bill independently per second, and billing stops when a sandbox stops; any credit is an account voucher described on the billing page 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 |
not stated |
| Binding | write-gated |
| Interface | unknown |
| Tooling | not stated |
| Composes with | not stated |
| Reliability | no reliability event recorded |
| Evidence rung | documented the pricing documentation was read on 2026-10-22 (Chinese). It defines billing per second on vCPU and RAM with billing stopped on sandbox stop, and states no free tier; vouchers are an account balance rather than a published allowance |
| Checked | 2026-10-23 (live); stale on 2026-11-22 |
Evidence (2)
| date | source | what it proved |
|---|---|---|
| 2026-10-02 | ppio.com/docs/sandbox/pricing.md | not published |
| 2026-10-22 | ppio.com/docs/sandbox/pricing.md | 'PPIO Sandbox 支持 vCPU 与内存(RAM)按秒独立计费' - vCPU and RAM bill separately per second - with '沙箱停止后不再计费' on both rows: billing stops when a sandbox stops. The page states no free tier and points at an account voucher balance |
More fields (15)
{
"talaria_m10h_usd": 19.31,
"talaria_m24h_usd": 46.35,
"nemo_m10h_usd": 19.31,
"keep_rate": 1.0,
"keep_rate_basis": null,
"billing_model": "unclassified",
"shape": null,
"session_cap": {
"wall_clock_min": null,
"idle_reap_s": null,
"restart_required": false,
"note": "billing stops when a sandbox stops, so stopping is the cost lever; the page states no automatic reap"
},
"persistent_storage": null,
"network_out": null,
"price_note": null,
"free_tier_kind": "none",
"burst_x": 1.0,
"ps": null,
"net": null
}Unknown or withheld values (17)
$.chainingnot-stated$.egress.meteredegress-unknown$.egress.termsnot-stated$.evidence[0].provedprocess$.extras.keep_rate_basisprocess$.extras.price_noteprocess$.identity_cost.detaildefaulted$.identity_cost.kindidentity-defaulted$.interfaceinterface-unknown$.interface_detailwithheld$.limits_notenot-stated$.per_object_cap.notesprocess$.per_object_cap.statednot-stated$.reliability_notenone-observed$.retention.notesnot-extracted$.retention.statednot-stated$.toolingnot-stated