Serverless Overview

Deploy and run your AI workloads on high-performance GPUs with Waverless, WaveSpeedAI’s serverless GPU platform.

What is Waverless?

Waverless is a serverless GPU task orchestration system designed for AI inference and training workloads. It provides on-demand access to powerful GPUs without managing infrastructure.

Key Features

Feature	Description
RunPod Compatible	Zero-code migration from RunPod with compatible API
Auto Scaling	Automatically adjusts worker count based on task queue depth
Multi-Endpoint	Isolate different applications through separate endpoints
Graceful Shutdown	Zero task loss during rolling updates and scale down
High Availability	Multi-replica deployment with no single point of failure

How It Works

1. Create Endpoint    →    Define your worker image and GPU spec
2. Deploy Workers     →    Workers auto-scale based on demand
3. Submit Tasks       →    Send tasks via API
4. Get Results        →    Receive results via polling or webhook

Use Cases

Custom Model Deployment — Run your own AI models on dedicated GPUs
Batch Processing — Process large volumes of data in parallel
Training Workloads — Fine-tune models with on-demand compute
High-Throughput Inference — Scale inference pipelines automatically

Architecture

Waverless uses a pull-based architecture where workers actively pull tasks from a queue:

Task Queue — Tasks are queued and distributed to available workers
Worker Pool — Workers pull tasks, execute them, and return results
Auto Scaler — Monitors queue depth and adjusts worker count

Getting Started

View GPU Pricing — See available GPUs and costs
Quick Start — Get up and running in minutes
Create Endpoint — Deploy your first endpoint
Build Worker — Write your handler code

Enterprise Access

Waverless is currently available for enterprise customers. To request access:

Go to wavespeed.ai/serverless
Fill out the request form
Our team will contact you to discuss your use case

Supported LLM Models GPU Pricing