#24 in Hosting & ComputeUpdated 2026-08

TensorPool

by TensorPool · On-demand multi-node GPU clusters managed from a CLI

39.8 — BenchRank score out of 100

Homepage
Pricing page

Visit website See full Hosting & Compute ranking

Overview

TensorPool provides on-demand multi-node GPU clusters driven from a CLI: 'tp cluster create' provisions nodes with 3.2Tb/s InfiniBand, 'tp storage create' adds shared volumes, and 'tp job push' starts training jobs through a Git-style interface. SSH access is also offered. Capacity is scheduled and binpacked onto clusters TensorPool operates with partners including Nebius, Lambda Labs and Google Cloud.

Best for: ML teams training foundation models who want on-demand multi-node GPU clusters driven from a CLI
Pricing: No platform fee is shown; GPUs are billed by the hour ($0.015/hr CPU, $1.49/hr L40S, $1.99/hr H100 SXM, $2.99/hr H200 SXM, $4.99/hr B200 SXM, $5.49/hr B300 SXM) with shared storage at $100/TB/month and object storage at $50/TB/month plus $0.005 per 1,000 requests, while Enterprise is contact-sales.
Runs on: Web

Strengths and trade-offs

Strengths

H100 SXM at $1.99/hr versus $3.29 quoted for Lambda Labs
Multi-node clusters on demand, always 3.2Tb/s InfiniBand
Shared storage up to 300 GB/s read and 150 GB/s write
Git-style CLI for training jobs; normal SSH access also offered

Trade-offs

Standard plan caps at 128 GPUs; beyond that needs Enterprise contact sales
99.9% uptime SLA, spend controls and monitoring are Enterprise-only
Capacity is binpacked onto partner providers, not TensorPool-owned hardware
Storage is billed separately at $100/TB/month for shared volumes

Pricing

Published plans from TensorPool’s own pricing page, in USD. Usage charges and add-ons may apply on top.

TensorPool pricing tiers, monthly rates in USD
Plan	Monthly	Includes
Standard Plan	—per GPU hour	On-demand access to up to 128 GPUs 24/7 support in TensorPool Slack High performance shared storage $20 free credit for starring the GitHub repo
Enterprise	Contact sales	Scale to thousands of GPUs Custom storage solutions, analytics and monitoring Spend controls and custom org features Dedicated support with 99.9% uptime SLA

How this score is made up

Each dimension is scored out of 100 and combined into the headline score using fixed weights.

MCP support: 70 out of 100
Documentation: 0 out of 100
Agent friendliness: 30 out of 100
Pricing transparency: 85 out of 100
Changelog: 0 out of 100
Marketing site structure: 65 out of 100
Operational trust: 0 out of 100

Measured, but not part of the score

Useful to know, but not a mark for or against the product — so these do not affect the ranking.

Openness: 60 out of 100
Maintenance: 65 out of 100

This doesn’t look right — report a problem with TensorPool’s score

Alternatives in Hosting & Compute

Ranked 1
79.2 — BenchRank score out of 100
Cloudflare
Cloudflare · Global network for CDN, security, serverless compute and Zero Trust access
Best for: Teams wanting CDN, DDoS protection, serverless compute and Zero Trust access on one network
Ranked 2
78.1 — BenchRank score out of 100
Docker
Docker · Container platform for building, sharing and running applications
Best for: Developer teams that build, share and run container applications and need admin and security controls
Ranked 3
69.8 — BenchRank score out of 100
Pulumi
Pulumi · Infrastructure as code in general-purpose programming languages
Best for: Platform teams defining multi-cloud infrastructure in general-purpose programming languages