# Arrakis vs fal.ai fal.ai is best known for hosted image, video and audio model APIs billed per output, plus serverless GPUs billed by the second. Arrakis runs the open models you choose on a GPU of your own, with your own engine options and root SSH, billed by the GPU-hour. ## 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 | fal.ai | |---|---|---| | B300 per GPU-hour | Live: [GPU prices](/gpus.html.md) | $12.99 list on Serverless (as low as $5.99) | | B200 per GPU-hour | Live: [GPU prices](/gpus.html.md) | $7.99 list on Serverless (as low as $5.49 with committed capacity) | | H100 per GPU-hour | Live: [GPU prices](/gpus.html.md) | $4.50 list on Serverless (as low as $2.49) | | Billing | By the hour; a started hour counts in full | Per second on Serverless; per output for model APIs | | How you pay | Prepaid balance: card, Link or Apple Pay | Prepaid credits | | Root SSH and any Docker image | Yes | Only on enterprise Compute machines | | Serverless endpoints, billed per request | No: each model gets a GPU of its own | Yes: fal Serverless, per second | | Hosted models, billed per output | No: deploy any Hugging Face model yourself | Yes: image, video, audio and language model APIs | | MCP server for AI agents | Yes, remote at /mcp | Yes: runs models, doesn't rent GPUs | | Spend cap on each API key | Yes, on every token | No caps per key; team spending alerts | | Automatic shutdown | Yes: every rental has a removal time | Yes: idle runners shut down | | Interruptible machines | Yes, where hosts offer them | No | | Searches several GPU markets | Yes | No: its own platform | 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 - **fal.ai fits better.** You want a hosted image or video model with nothing to run, billed per output, or a large catalog of ready-made generative APIs. - **Arrakis fits better.** Your GPU is busy for hours, you want your own model version and engine options, root access, or one hourly price however many pictures you make. ## 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 **When is a GPU of my own cheaper than fal.ai?** Once it is busy. On October 31, 2026, fal.ai listed an H100 at $4.50 an hour on its Serverless platform and billed most hosted models per output. A GPU on Arrakis costs its hourly price however many pictures it makes, so steady generation favors your own GPU and a few pictures a day favor paying per output. **Can I run the same models as on fal.ai?** The open ones, yes: FLUX, Qwen-Image, Wan and other open-weight models run on vLLM-Omni or SGLang on Arrakis. Closed models that fal.ai hosts for their makers can't be downloaded. ## Sources - - - - - -