AutoDL review
Updated September 28, 2026 · DeployCue Editorial Team · Facts checked September 26, 2026
We earn a commission when you sign up through links on this site. Partner status can affect where a brand appears in our rankings.
Our verdict
AutoDL suits developers and small teams who need a broad GPU range on hourly terms, since instances bill at per second granularity. Bandwidth and traffic are not billed separately, so short experiments stay cheap, and new accounts get Registration gift: 炼丹会员 membership valid 1 month.
Key facts
- Best for
- Short GPU rentals billed per second, with no separate charge for bandwidth or traffic.
- GPU models
- Tesla P40, TITAN Xp, GTX 1080 Ti, RTX 2080Ti, Tesla V100, RTX 3060, RTX A4000, RTX 3080Ti, RTX A5000, RTX 3090, NVIDIA A40, A100 SXM4, RTX 4090, RTX 4090D, A800, L20, H20, H800, Cambricon MLU 200 series, Huawei Ascend NPU, Moore Threads GPU (Source, checked September 26, 2026)
- Billing granularity
- Per second (Source, checked September 26, 2026)
- Egress fee
- No separate charge for instance bandwidth or traffic (Source, checked September 26, 2026)
- Free credits
- Registration gift: 炼丹会员 membership valid 1 month (Source, checked September 26, 2026)
- Interconnect
- NVLink within A100 SXM4 and H20-NVLink zones; multi-node multi-GPU parallelism documented (Source, checked September 26, 2026)
- Legal entity
- SeetaCloud Technology Co., Ltd (Zhongke Shituo (Nanjing) Technology Co., Ltd) (Source, checked September 26, 2026)
- Headquarters
- China (Source, checked September 26, 2026)
- Support channels
- Chat, Help-centre, Community (Source, checked September 26, 2026)
- Pricing model
- Hourly billing
- Official website
- autodl.com
Pros and cons
Pros
- Instances bill at per second granularity, so a short job stops costing the moment you release it.
- No separate charge for instance bandwidth or traffic, so checkpoints, images and datasets move in and out for nothing extra.
- The Tesla P40, TITAN Xp and GTX 1080 Ti catalogue pairs older consumer cards with current data centre parts.
- NVLink inside A100 SXM4 and H20 NVLink zones, with multi-node multi-GPU parallelism documented.
Cons
- Support runs on chat, help centre and community only, so a phone line or email route is not listed.
- The listed billing model is hourly, so there is no monthly or reserved rate to compare.
- The provider sits in CN and is run by SeetaCloud Technology Co., Ltd, so buyers elsewhere should settle payment terms first.
Who it suits
- Developers who stop a run the moment it finishes, since billing is per second.
- Teams that want non-NVIDIA options, since the catalogue also lists Cambricon, Huawei Ascend and Moore Threads parts.
- Solo learners who put Registration gift: 炼丹会员 membership valid 1 month to work before regular spend.
Who should look elsewhere
- SberCloud, if you want a mainstream provider whose GPU line is B300, H100 and A100.
- ComputePrices.com, if you want a single page tracking H100 SXM, A100 SXM and H200.
- Salad, if your jobs run as containers on distributed hardware with NVIDIA RTX 3060, NVIDIA RTX 3070 and NVIDIA RTX 3070 Ti.
How AutoDL compares
Where AutoDL is available
- Available
- China
Countries from the brand (Source, checked September 26, 2026)
Frequently asked questions
How much does a GPU hour cost on AutoDL?+
AutoDL rents instances hourly, so you pay for the cards you run. Across the listings compared here the median entry rate is €0.63, with a spread of €0.39 to €1.75. Check the live rate before a long run.
What happens if I stop a training job early?+
Billing is per second, so an instance only costs for the time it runs. That suits experiments you may abandon after a few steps, and work where a long commitment is the risk.
Is there an extra charge for bandwidth or pulling datasets?+
No separate charge for instance bandwidth or traffic, so downloading checkpoints, images and datasets adds no line item. That keeps estimates simple for teams moving large artefacts in and out of the platform, because the card rate stays the main number.
Which GPUs can I rent?+
The catalogue covers Tesla P40, TITAN Xp and GTX 1080 Ti. Consumer and professional cards sit next to current data centre parts, and non-NVIDIA hardware is listed too, so a workload without an NVIDIA card may still have somewhere to go.
Can I link several GPUs for a large training run?+
NVLink is available inside A100 SXM4 and H20 NVLink zones, and multi-node multi-GPU parallelism is documented, so larger jobs are not limited to a single card. Billing stays per second on a multi-node run, and the listed model is hourly.
Is there a free trial or starting credit?+
Signing up includes Registration gift: 炼丹会员 membership valid 1 month, enough to try a card before you spend real money. Read it as a short sampler rather than a standing free tier, and watch usage while the membership window is open.
How do I get help if something goes wrong?+
Support runs on chat, help centre and community. That suits developers comfortable troubleshooting in a chat window or a community thread, and self-serve work better than teams wanting a phone line or a named contact on a contract.
Which country is AutoDL based in, and who runs it?+
The provider is headquartered in China and is run by SeetaCloud Technology Co., Ltd. That shapes where machines sit and how billing is handled, so check payment and invoicing terms early if you buy from abroad.
Sources
- autodl.com/docs checked September 26, 2026
Backs: Free credits, Support channels
- autodl.com/docs/gpu checked September 26, 2026
Backs: GPU models, Interconnect
- autodl.com/docs/network checked September 26, 2026
Backs: Egress fee
- autodl.com/docs/price checked September 26, 2026
Backs: Billing granularity, Headquarters, Legal entity, Pricing model
“户名：中科视拓（南京）科技有限公司 ... Beneficary Name: SEETACLOUD TECHNOLOGY CO.,LTD”
Compiled from 4 source pages, checked September 26, 2026. We did not open an account.
Not confirmed yet: GPU hour from.
AutoDL
Short GPU rentals billed per second, with no separate charge for bandwidth or traffic.