TensorPool

TensorPool

TensorPool provides on-demand multi-node GPU clusters for training foundation models without managing infrastructure.

PaidWebOpen source
Visit websitetensorpool.dev
Compare with DataRobotExplore TensorPool alternatives

Overview

TensorPool provides on-demand multi-node GPU clusters for training foundation models without managing infrastructure. It features high-throughput storage optimized for distributed training and CLI-based orchestration to eliminate idle GPU time via a push/pull paradigm. Targeting ML engineers and researchers, it offers cost-effective access to H100, H200, and B200 GPUs. Pricing starts at $1.49/hr for L40S GPUs, with $20 free credits available for GitHub stars.

Key features

  • On-demand multi-node GPU clusters
  • High-throughput distributed training storage
  • CLI-based orchestration with push/pull paradigm

Pros & cons

Collected from user feedback found in web search

Pros

  • Command-line interface for GPU resource management
  • Supports multiple cloud providers
  • Founded by Stanford graduates with ML expertise
  • Y Combinator backed

Cons

  • Extremely early stage (founded 2025)
  • Limited adoption data

Pricing

PaidStarting price: US$1.49/mo
View pricing page

Verified on: · Recheck pricing on the official page

Use cases

  • Training foundation models
  • Distributed AI workload management
  • High-performance data storage for ML

Who it is for

ML engineersResearchers

Integrations

AWSGoogle CloudMicrosoft AzureLambda LabsNebiusLightning AIGitHubSlack

Tags

API

How we verified this

Company, pricing, and feature details come from the primary sources below and our latest verification pass. When sources disagree, the official source and the most recent check win.

Last verified 08/02/2026Verified sources: 1

Alternatives

Tools you can use instead