Release2026-08-04

Crusoe made its Serverless Fine-Tuning service generally available in its Intelligence Foundry platform, letting customers adapt open language models without provisioning their own graphics hardware. Billing is per unit of training text processed, and jobs stop automatically once the model stops improving. The service runs a fixed low-rank adaptation workflow, so teams needing fully custom training code must use ordinary Crusoe Cloud instances instead.

What changed

Fine-tuning open models on Crusoe required renting and managing GPU instances billed by the hour.

What it unlocks

Running a managed fine-tuning job on an open model from a browser, software kit or API, then deploying it in one click or downloading the weights to run elsewhere.

  • 12 open models available at launch across 6 providers
  • LoRA fine-tuning billed per training token rather than per GPU-hour

What you need to act on it

  • a Crusoe Cloud account with access to Intelligence Foundry

Send this to someone who needs it

Shares the story and its sources. Nothing about you.

What does this mean for your job?

This is the story as everyone gets it. Once a week we send you the version written for your role — what changed, why it matters for the work you actually do, and one thing to try. Free while we tune it.