# Arrakis vs Vast.ai Vast.ai is a marketplace where independent hosts rent out GPUs at prices set by supply and demand, billed by the second. Arrakis searches several independent GPU markets at once, rents on one prepaid balance and gives every agent token a spend cap and every rental a removal time. ## Side by side As of 31 Oct 2026, 17:40 UTC, the lowest price for the NVIDIA B300 on Arrakis is $4.60 per GPU-hour, with 73 GPUs available on 10 machines. | | Arrakis | Vast.ai | |---|---|---| | B300 per GPU-hour | Live: [GPU prices](/gpus.html.md) | Varies: from $9.94, median $15.00 | | B200 per GPU-hour | Live: [GPU prices](/gpus.html.md) | Varies: from $6.25, median $9.06 | | H100 per GPU-hour | Live: [GPU prices](/gpus.html.md) | Varies: H100 SXM from $1.73, median $2.40 | | Billing | By the hour; a started hour counts in full | Per second | | How you pay | Prepaid balance: card, Link or Apple Pay | Prepaid credits | | Root SSH and any Docker image | Yes | Yes, as root in a container | | Serverless endpoints, billed per request | No: each model gets a GPU of its own | Yes: Vast Serverless, per second of worker time | | Hosted models, billed per output | No: deploy any Hugging Face model yourself | No priced model catalog found | | MCP server for AI agents | Yes, remote at /mcp | No official server: an agent skill and plugins | | Spend cap on each API key | Yes, on every token | No spend limit per key | | Automatic shutdown | Yes: every rental has a removal time | No timer; an instance can stop itself | | Interruptible machines | Yes, where hosts offer them | Yes: interruptible machines by bidding | | Searches several GPU markets | Yes | One marketplace of hosts | Facts about other providers come from their own pages, checked on October 31, 2026. Prices and features change: check them before you choose. Arrakis is not affiliated with them. ## Which fits you - **Vast.ai fits better.** You want billing by the second, the cheapest interruptible machines by bidding, or the widest choice inside one marketplace. - **Arrakis fits better.** You want several markets searched at once instead of one, hard spend caps and removal times for agents, an MCP server built in, or a model served with your own options in one step. ## Moving to Arrakis Your Docker image, SSH keys and scripts move as they are: Arrakis boots any public or private image with root SSH, and serves your own model with your own engine options. Start with one machine in the console, or point your agent at /mcp with a capped token. ## Questions **Is Arrakis cheaper than Vast.ai?** It depends on the hour. On October 31, 2026, Vast.ai listed H100 SXM machines from $1.73 an hour, with a median of $2.40, plus storage and bandwidth at rates each host sets. Arrakis shows the lowest live price across the markets it connects. Compare both when you rent. **Does Vast.ai have an MCP server?** Not an official one on October 31, 2026: Vast.ai publishes an agent skill and plugins that wrap its CLI. Arrakis runs a remote MCP server at /mcp, with a spend cap on every token. ## Sources - - - - - - - - -