Skip to content

DirectoryTools

Float16.cloud logo

Float16.cloud

GPU infrastructure on credit-based quotas.

Visit Float16.cloud

Coding & WebFreemiumWe checked the prices on float16.cloud on 28 Sep 2026.

Float16.cloud homepage screenshot

What is Float16.cloud?

GPU infrastructure on credit-based quotas.

Who is Float16.cloud best for?

Float16.cloud is best for teams that share a small pool of GPUs and want quotas instead of fixed time slots.

  • Teams sharing a few GPUs

    Credit-based quotas replace fixed time slots, so each team spends its GPU hours when its work needs them rather than in a booked window.

  • Several models on one card

    NVIDIA MIG splits one GPU into as many as 7 isolated instances, each with its own memory.

  • Developers who need an LLM endpoint

    One-click deploys put models such as Qwen 3, Typhoon and GPT-OSS behind an OpenAI-compatible API without setting up servers.

  • Research groups

    Jupyter and SSH access come with ready templates for Parabricks, Clara, MONAI and AlphaFold.

How much does Float16.cloud cost?

Float16.cloud charges $4.32 an hour for an on-demand H100 GPU and $2.16 an hour for a spot one, paid per second from prepaid credit, while its self-hosted Starter tier for up to 5 GPUs is free.

Starter$0Free forever
  • Self-hosted GPU management for teams up to 5 GPUs
  • GPU passthrough VMs and NVIDIA MIG; no spot VMs or one-click deploy
  • Set up through a demo request
H100 Spot$2.16Per GPU hour
  • Half the on-demand rate, for work that can be interrupted
  • Billed per second from prepaid credit
  • Storage $1.00 per GB per month
H100 On-Demand$4.32Per GPU hour
  • Dedicated H100 with SSH, Jupyter and Docker
  • Billed per second from prepaid credit
  • A stopped instance pays only for storage
Serverless H100$4.32Per hour ($0.0012 per second)
  • Scales to zero when idle
  • Spot rate $2.16 per hour ($0.0006 per second)
  • Ready-made templates for models such as Qwen3

Float16 Cloud runs on prepaid credit held in US dollars: top up by card through Stripe or by bank transfer, and running instances stop when the balance reaches zero. H100 instances run in Thailand. The self-hosted Scale tier (up to 50 GPUs) and Hyperscale tier (unlimited GPUs) are priced on request. LLM as a Service sells a dedicated H100 endpoint by the month, but its page does not show the price. Full pricing on float16.cloud. We checked these prices on .

How do readers rate Float16.cloud?

No one has rated Float16.cloud yet. Be the first.

No ratings yetThe score and its stars show once 3 readers have rated Float16.cloud.
Your rating

Sign in to rate. One rating per account.

Your rating of Float16.cloud

What are the best Float16.cloud alternatives?

Three tools on Recaply do Float16.cloud's job.

Questions about Float16.cloud

What is Float16.cloud?

GPU infrastructure on credit-based quotas.

Who is Float16.cloud for?

Float16.cloud is best for teams that share a small pool of GPUs and want quotas instead of fixed time slots.

Is Float16.cloud free?

Partly. Float16's self-hosted Starter tier is free forever for teams with up to 5 GPUs, though you get it through a demo request. GPUs on Float16 Cloud are paid per second from prepaid credit, from $2.16 an hour for a spot H100.

How much does Float16.cloud cost?

Float16.cloud's paid plans are H100 Spot at $2.16 per GPU hour, H100 On-Demand at $4.32 per GPU hour and Serverless H100 at $4.32 per hour ($0.0012 per second).

What are the best Float16.cloud alternatives?

The closest alternatives on Recaply are CLORE.AI, Salad and Modal.

Where can I try Float16.cloud?

On its official website, float16.cloud.

How does Float16 bill GPU time?

Float16 Cloud uses prepaid credit held in US dollars. You top up by card or bank transfer, and GPU time is deducted per second while instances run. Running instances stop when the balance reaches zero, and stopped instances pay only for storage at $1.00 per GB a month.