Modal

Serverless Python for GPU workloads.

Freemium 16–80 GB VRAMSelf-hosted server

Actually FreeWatermark-FreeHobbyist-FriendlyAPI

Visit ModalUpdated 2026-05-09 · Direct link

Hardware requirements

Self-hosted server · Datacenter GPU (80 GB+)

16–80 GB VRAM

Min VRAM

16 GB

Rec. VRAM

80 GB

Min RAM

16 GB

Rec. RAM

128 GB

Disk

100 GB

GPU class

Datacenter GPU

Provided by their imagesNo Apple SiliconCPU-Capable

T4 → H100 available. B200 in preview.

Screenshot placeholder · Modal

What is Modal?

Define functions in Python; Modal runs them on-demand on GPUs. Excellent for batched ComfyUI workflows, fine-tunes, and exposing models as HTTPS endpoints with autoscaling.

Pros & cons

Pros

Cleanest dev experience for serverless GPU
Autoscaling to zero
Good cold-start mitigations

Cons

Locked to their SDK
Less raw control than a VM

What's actually free?

$30/month free compute credit on signup.

✓ Actually FreeWatermark-Free

Alternatives

RunPod

On-demand GPU pods for ComfyUI, vLLM, training.

Paid 8–80 GB VRAM

Min VRAM: 8 GB
GPU class: Datacenter GPU
Quant: FP16

Watermark-FreeHobbyist-FriendlyAPI

Replicate

Run any open-source model with one API call.

Paid☁ Cloud · no GPU

Watermark-FreeHobbyist-FriendlyAPI

Beam

Serverless GPU functions — deploy a Python file, get an HTTPS endpoint.

Paid☁ Cloud · no GPU

Watermark-FreeHobbyist-FriendlyAPI