Hermes VPS pricing: choose your plan.
- Operating system
- Linux· SSH
- Location
- New York, USA
No contracts, cancel anytime.
Early Black Friday deal· 60% off your first month
Run Hermes Agent from Nous Research on a dedicated Linux VPS.
HermesLinux VPS · New YorkRunningyou@vps:~$hermes
›Every weekday at 7 am, send me a research brief on Telegram.
›
Explore AI tools for your cloud VPS
1 of 6Choose your Hermes VPS
No contracts, cancel anytime.
Plan selection
Size for everything you run at once
Specifications
Compare PlansServer Specifications
Agents run on your plan's CPU, memory and storage with your own API keys. Local models need more memory.
Unsure which agent to use? Our team helps you set it up at no charge.
Server Specifications
Server Specifications
Server Specifications
Server Specifications
Specifications
Compare PlansServer Specifications
Server Specifications
Server Specifications
Server Specifications
Server Specifications
Included at no extra cost. Hover or tap a feature for details.
The details behind every value on the plan cards. Hover or tap a value on a card for the same details.
High-Frequency plans are 85% faster per thread.
Memory for the agents, models and services you run at the same time.
Leave room for builds, tests and additional workspaces.
Room for the operating system, your agents, tools and files.
The operating system and installed software use part of it. Choose a larger plan to keep more data.
Choose the system your software needs at checkout.
Either system lets you install compatible runtimes, tools and services.
Speed is how much data moves at once.
Actual throughput depends on the network route and workload.
The typical time from purchase until your VPS is ready.
An extra recovery point for your VPS, at no extra cost. RAID1 mirrors data across drives to protect against a single drive failure.
Keep separate copies of important files as well. RAID1 is not a backup. It does not protect against deletion or data corruption.
Agents you can install: Claude Code (Linux), Codex (Linux), OpenCode (Linux), Hermes (Linux), OpenClaw (Linux), Cursor CLI (Windows or Linux), AutoGPT (Linux), n8n (Linux), LM Studio (Windows), Gemini CLI (Linux).
Agents run on your plan's CPU, memory and storage with your own API keys. Local models need more memory. Your own MCP servers and the QuantVPS REST API work too. Unsure which agent to use? Our team helps you set it up at no charge.
| Plan | CPU Mark | Single-thread | Processor |
|---|---|---|---|
| VPS Starter | 2,550 | 2,554 | AMD EPYC 7003 |
| VPS Lite | 5,092 | 2,554 | AMD EPYC 7003 |
| VPS Pro | 8,020 | 2,554 | AMD EPYC 7003 |
| VPS Ultra | 15,185 | 2,554 | AMD EPYC 7003 |
| VPS Starter+ | 8,100 | 4,728 | AMD Ryzen 9 9950X |
| VPS Lite+ | 15,600 | 4,728 | AMD Ryzen 9 9950X |
| VPS Pro+ | 22,200 | 4,728 | AMD Ryzen 9 9950X |
| VPS Ultra+ | 28,100 | 4,728 | AMD Ryzen 9 9950X |
| Dedicated | 45,247 | 3,477 | AMD Ryzen 9 5950X |
| Dedicated+ | 65,708 | 4,728 | AMD Ryzen 9 9950X |
Logarithmic scale from 1,000 to 65,708. Bar lengths are not proportional to scores. Derived from PassMark CPU Mark (cpubenchmark.net, data through Oct 5, 2026) for the installed processor, scaled to each plan's vCPUs. Dedicated scores are bare-metal averages, from PassMark CPU Mark (cpubenchmark.net, data through Oct 5, 2026). Not a live test of your VPS or server. Some Dedicated+ servers run the Ryzen 9 9900X: 54,290 CPU Mark, 4,670 single-thread.
No hidden fees, no exit fees, rate locked for at least three years, and a 14-day money-back guarantee.
Prorated Billing
There is no exit fee. You pay only the difference, and unused time on your current plan is automatically credited toward the upgrade.
Upgrade in Minutes
Process the upgrade from your dashboard with no ticket or waiting. The additional resources take effect after a brief reboot.
Keep Your Setup
Your files, settings, and tools stay exactly as they are when you move to a more powerful plan.
Plan Help Before Checkout
Use the configurator or ask support to map your tools, location, and workload to the right plan.
Tell us what you run. Get the plan that fits.
We size the CPU, memory, and storage around the agents and tools you actually run, with no guesswork and no upsell to a tier you will not use.
2 of 6How it works
01Secure remote accessSSH from your preferred terminal
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash. Hermes needs Python 3.11 or newer.
Workflow stage
The complete workflow behind this stage
02Continuous operationUbuntu 24.04
hermes setup, then Nous Portal, OpenRouter, Anthropic, OpenAI or a local model server. Pick a model with at least 64K tokens of context.
Workflow stage
The complete workflow behind this stage
03New York, USARunning 24/7 on dedicated resources
hermes gateway install plus sudo loginctl enable-linger $USER runs the gateway as a service that survives logout and reboots. You talk to it from Telegram, Discord, Slack, WhatsApp or Signal.
Workflow stage
The complete workflow behind this stage
3 of 6Why Hermes on a VPS
4 of 6Who runs Hermes
A day on your VPS
4 workloads. One server, all day.
Each stays available on your VPS whether or not you are connected.
5 of 6Uptime & security
99.999% monthly VPS availability, backed by an industry-leading uptime guarantee SLA: a service credit for any covered outage of a minute or more. Availability and credits are assessed monthly under the SLA’s measurement rules, coverage and exclusions.
Service availability
Read the commitment alongside the history
Core infrastructure. Redundant systems ensure continuous operation.
Network availability. Multiple carriers and automatic failover.
SLA-backed promise. Contractual uptime guarantee with credits.
First service creditMinutes down in a 30-day month
Competitors A to D: from each provider’s own published SLA or terms, read on 4 Nov 2026. Their terms can change. Credits are requested within 30 days after the month ends.
Server & network availability · 99.999% SLA
Tier IV.
Concurrently maintainable facilities. Built to keep your agents and services running.
Explore infrastructureRedundant by design.
Tier IV facilities run N+1 power and cooling. The network fails over between tier-1 carriers by BGP. When one side goes down for maintenance or a fault, the other side carries your server.
24/7
A NOC team monitors every server, network path, and facility signal around the clock.
Meet the NOC teamFiltered at the edge.
Always-on mitigation filters attack traffic at the network edge. Every plan includes a dedicated IP and unmetered bandwidth for your workloads.
6 of 6Questions, answered