VESSL Cloud

What it does

VESSL Cloud takes in a request for GPU capacity and returns running compute: persistent single-node workspaces with SSH and IDE access, one-command batch jobs that queue for GPUs and stream logs and metrics, multi-node InfiniBand clusters with root access, and storage shared across workspaces and jobs. For inference, it offers Provisioned Throughput, where you contract throughput in PTUs and send requests to an endpoint compatible with the OpenAI and Anthropic APIs.

People use it to prototype in a workspace on their own GPU and then submit the same environment as a job, without rebuilding it. The vesslctl command line tool lets a coding agent such as Claude Code run the same GPU commands. Self-serve on-demand GPUs include H100, A100, and L40S, billed by the second; H200 and Blackwell GPUs are reserved only, available through the sales team.

It suits AI research teams, engineers training or fine-tuning models, and organizations that want managed throughput for open models. Access requires a VESSL Cloud account, and Provisioned Throughput requires a sales-arranged contract. Limitations: managed Kubernetes and Slurm are not yet available, and reserved pricing is not published.

What it can do

• Launch single-node GPU workspaces with up to 8 GPUs and SSH access from your own IDE
• Submit one-command batch jobs for training, fine-tuning, and batch inference with live logs and metrics
• Run multi-node dedicated clusters over 400 Gbps InfiniBand with full root access
• Serve open models through Provisioned Throughput with OpenAI- and Anthropic-compatible APIs
• Drive GPU provisioning from a terminal or coding agent with the vesslctl CLI
• Attach cluster storage and object storage to workloads

Works with

• OpenAI Chat Completions API compatibility
• Anthropic Messages API compatibility

Before you choose

• Only H100, A100, and L40S are self-serve; H200 and Blackwell GPUs need a reserved contract
• Provisioned Throughput is priced per PTU and arranged with sales; the $0.05 per PTU per minute rate is indicative and subject to change
• Multi-node training requires a dedicated VM Cluster; single workspaces and jobs are limited to one node
• Managed Kubernetes and Slurm are on the roadmap and not yet available

Listed under

• Coding
• Sandboxes and infrastructure
• AI infrastructure
• Cloud compute
• GPU cloud
• Inference
• ML teams
• Model training

Get started

Self-serve on-demand GPUs start from the sign-in page; inference and reserved capacity start with a sales consultation.

1. Sign in to cloud.vessl.ai and choose an on-demand GPU such as H100, A100, or L40S
2. Install vesslctl with the install script at api.cloud.vessl.ai and run vesslctl skill install if you use a coding agent
3. For Provisioned Throughput, consult sales, agree the PTU count and terms, then set the base URL and API key in your existing client

On-demand GPU compute is billed by the second. Storage is billed daily on stored data. Provisioned Throughput is billed monthly per PTU. Reserved and H200/Blackwell pricing is quoted by term and volume.

Source: vessl.ai/en/pricing, checked 29 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• Container Compute: GPU workspaces and one-command batch jobs | VESSL Cloud ↗vessl.ai
• Provisioned Throughput: guaranteed throughput for open models | VESSL Cloud ↗vessl.ai
• VESSL Cloud | The AI Cloud for All Models ↗vessl.ai
• Pricing | VESSL Cloud ↗vessl.ai
• VESSL Cloud Documentation ↗docs.cloud.vessl.ai

Company website

vessl.ai: product, pricing and policy pages

YC directory mirror

Batch and company name

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

manual_review

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.