按功能
- 可用 12 指南
- 打开指南查看匹配的提供商
- 使用提供商卡上的比较功能建立候选列表
提供API和CLI管理的云GPU服务商
An API or CLI interface allows you to programmatically provision, manage, and tear down GPU instances — essential for MLOps...
支持 Docker 和自定义镜像的云 GPU 提供商
Docker support allows you to bring your own environment with pre-installed frameworks, CUDA versions, and dependencies, ensuring reproducibility between development...
支持 Jupyter Notebook 的云 GPU 提供商
Jupyter Notebooks provide an interactive development environment that is widely used in data science and ML research. Having Jupyter pre-configured...
支持 Kubernetes 的云 GPU 提供商
Kubernetes has become the standard for orchestrating ML training and inference workloads at scale. GPU-aware Kubernetes clusters enable automated scheduling,...
支持多节点GPU集群的云GPU提供商
Training models that exceed the memory capacity of a single node requires multi-node GPU clusters with fast inter-node networking. Multi-node...
具有 NVLink 或 InfiniBand 的云 GPU 提供商
High-bandwidth GPU interconnects like NVLink (up to 900 GB/s) and InfiniBand (up to 400 Gb/s) are essential for efficient multi-GPU...
提供按秒计费的云GPU服务商
Per-second billing ensures you pay only for the exact compute time consumed, which is particularly valuable for short experiments, iterative...
具有持久存储的云GPU提供商
Persistent storage ensures that your datasets, model checkpoints, and training outputs survive instance restarts and shutdowns. Without persistent storage, you...
具备无服务器GPU推理的云GPU提供商
Serverless GPU eliminates idle costs by automatically scaling your inference endpoints to zero when not in use, and spinning up...
提供抢占式/可中断实例的云GPU供应商
Spot or preemptible GPU instances offer 50-90% savings compared to on-demand pricing, in exchange for the possibility of interruption during...
提供 SSH 访问的云 GPU 服务商
SSH access gives you full root-level control over your GPU instance, allowing you to install custom software, debug issues, manage...
零出站费用的云GPU提供商
Egress fees — charges for transferring data out of the cloud — can add significant unexpected costs when exporting model...