Packet AI
- Free allowance
- No free GPU tier is published; the rate card read on 2026-11-03 shows A100 80GB at USD 0.89 an hour on the Dynamic plan and USD 1.43 on the Dedicated plan, L40S 48GB at USD 0.92 on the Dedicated plan, RTX 6000 Pro at USD 0.66 and B200 at USD 3.75 Dynamic and USD 6.99 Dedicated.
- Verdict
- paidno free allowance; listed for its price
- Cost
- no price stated
- Identity
- cardcard required, no phone
- Binding limit
- write-gateda launch needs a funded wallet: the minimum to launch is the first 30 minutes at the pod's hourly rate; Premium GPUs unlock after the first paid top-up and a voucher alone does not launch them
- Evidence
- documented28 sources; a first-party page was read and the sentence quoted
- Freshness
- stale on 2026-12-04checked 2026-11-03 (live); stale on 2026-12-04
| Free tier | no free GPU tier is published; the rate card read on 2026-11-03 shows A100 80GB at USD 0.89 an hour on the Dynamic plan and USD 1.43 on the Dedicated plan, L40S 48GB at USD 0.92 on the Dedicated plan, RTX 6000 Pro at USD 0.66 and B200 at USD 3.75 Dynamic and USD 6.99 Dedicated. The RTX 4090 row shows no on-demand price on the rate card; the comparison table lists 'RTX 4090' at USD 0.39 as the packet.ai starting rate. H100 SXM and H200 are labelled 'Coming soon'. No L4 is listed. unit none; metered on wall-clock; Dynamic is a shared card with scheduler-enforced isolation; Dedicated is a whole card; hourly pods are billed in 30-minute periods from a prepaid wallet |
|---|---|
| Price | none stated |
| Limits |
A pod launch is gated on a funded wallet (USD 5 to USD 10,000 per top-up by Stripe Checkout); there is no free allowance. |
| Identity | card Account creation is by GitHub, Google or email and the sign-up page says 'No credit card needed'; launching a pod needs a wallet top-up of USD 5 to USD 10,000 through Stripe Checkout, which saves the card for auto top-up. No phone or ID document is stated on the pages read. |
| Egress | not published on the pages read |
| Interface | rest, cli, ssh taken from the documentation index, which lists a REST API reference, a CLI and browser or SSH terminal access; not probed |
| Tooling | REST API with an API key, a CLI, SSH and a browser terminal, Jupyter and VS Code Server one-click apps, vLLM model serving, persistent storage volumes |
| Composes with | Hourly A100 80GB and L40S at USD 0.89 to USD 1.43 and USD 0.92 with SSH and an API, plus a token-billed serverless catalog drawing on the same wallet. |
| Reliability | The aggregator profile states B200s sold out and H200s coming soon; the provider is listed as newer with limited locations. The rate card's H100 SXM and H200 rows read 'Coming soon' on 2026-11-03. |
| Notes | Found via computeprices.com (provider slug packet-ai). The aggregator shows A100 SXM at USD 0.890, L40S at USD 0.920 and RTX 4090 at USD 0.390 in US East Virginia; the provider rate card places the A100 80GB in EU Frankfurt. Not read: the RTX 4090 on-demand rate and region. |
| Parsed numbers |
|
Evidence 28
| date | source | what it proved |
|---|---|---|
| 2026-11-03 | packet.ai/pricing | A100 80GB Dynamic plan rate is USD 0.89 an hour |
| 2026-11-03 | packet.ai/pricing | A100 80GB Dedicated plan rate is USD 1.43 an hour |
| 2026-11-03 | packet.ai/pricing | L40S Dedicated plan rate is USD 0.92 an hour |
| 2026-11-03 | packet.ai/pricing | the comparison table lists RTX 4090 with a packet.ai starting rate of USD 0.39 |
| 2026-11-03 | packet.ai/pricing | H100 SXM is labelled Coming soon |
| 2026-11-03 | dash.packet.ai/docs/billing | wallet top-ups are USD 5 to USD 10,000 through Stripe Checkout |
| 2026-11-03 | dash.packet.ai/docs/getting-started | a funded wallet is required before a pod launches |
| 2026-11-03 | dash.packet.ai/account | the account page offers sign-up without a card |
| 2026-11-03 | computeprices.com/providers/packet-ai | listed on the computeprices index |
| 2026-11-03 | packet.ai/pricing | the L40S Dedicated rate of USD 0.92 an hour on the rate card |
| 2026-11-03 | packet.ai/pricing | the A100 80GB Dedicated rate of USD 1.43 per GPU-hour and the Dynamic rate of USD 0.89 on the rate card |
| 2026-11-03 | packet.ai/pricing | the A100 80GB Dynamic rate of USD 0.89 an hour |
| 2026-11-03 | packet.ai/pricing | the published card set: an H100 SXM and an H200 are not yet orderable, and no L4 is listed |
| 2026-11-03 | packet.ai/pricing | no ingress and no egress fee, with local NVMe storage included up to 2 TB per node |
| 2026-11-03 | packet.ai/dedicated-gpu-cloud | the RTX 4090 Dedicated starting rate of USD 0.39 per GPU-hour, the cheapest Dedicated card |
| 2026-11-03 | packet.ai/dynamic-gpu-cloud | the Dynamic page marks the RTX 4090 as not yet orderable |
| 2026-11-03 | packet.ai/ | the home page lists an H100 at USD 2.50 an hour with a waitlist link rather than a deploy link |
| 2026-11-03 | dash.packet.ai/account | the account form says no credit card is needed to create the account and offers an email sign-up |
| 2026-11-03 | dash.packet.ai/docs/billing | GPU pods bill from a prepaid wallet in 30-minute periods |
| 2026-11-03 | dash.packet.ai/docs/billing | wallet top-ups run from USD 5 to USD 10,000 through Stripe Checkout |
| 2026-11-03 | dash.packet.ai/docs/billing | the card used to top up is saved |
| 2026-11-03 | dash.packet.ai/docs/billing | a launch is refused when the wallet cannot cover the first 30 minutes |
| 2026-11-03 | dash.packet.ai/docs/billing | a stopped pod is billed at the full hourly rate |
| 2026-11-03 | dash.packet.ai/docs/billing | part of the text of Limits, above |
| 2026-11-03 | dash.packet.ai/docs/billing | shared volumes of an account with a negative balance are scheduled for deletion after a 3-day grace period |
| 2026-11-03 | dash.packet.ai/docs/getting-started | pods authenticate with a password over SSH by default |
| 2026-11-03 | dash.packet.ai/docs | a REST API authorised with an API key |
| 2026-11-03 | gpuperhour.com/providers/packetai | a price tracker lists the provider with 43 offers, RTX 4090 at USD 0.39 per GPU-hour and every offer shown Sold Out on that date |
Unknown values 3
- egress metered: egress metering was not published
- per object cap stated: the field does not apply to this kind of offer
- retention stated: the field does not apply to this kind of offer