Modal

Modal is a serverless cloud platform designed to run compute-intensive AI and machine learning applications without requiring users to manage infrastructure. It provides an AI-native runtime with fast model initialization and autoscaling, a built-in globally distributed storage layer for high-throughput model and data access, and first-class support for GPU and CPU workloads for inference, training, and batch processing. Modal exposes APIs and SDKs to deploy OpenAI-compatible model endpoints, run notebooks and sandboxes, and orchestrate large-scale batch jobs.

• AI-native runtime — a serverless runtime engineered for heavy AI workloads with fast startup and autoscaling to handle parallel GPU/CPU jobs.
• Built-in distributed storage — globally distributed storage optimized for fast model loading and dataset access to reduce I/O bottlenecks.
• Flexible compute options — per-second GPU and CPU pricing for many GPU types and hardware configurations, enabling fine-grained usage-based billing.
• Model serving & inference — deploy OpenAI-compatible LLM endpoints and host custom model APIs at scale.
• Batch, notebooks, and sandboxes — run large-scale batch workflows, interactive notebooks, and isolated sandboxes for development and testing.

To get started, sign up on the web console, consult the documentation for SDK and API examples, and deploy a sample model or notebook to validate autoscaling and compute billing.