C

Serverless Fine-Tuning

API service

by Crusoe

Managed LoRA-based fine-tuning of open-weight models, billed per token of training data rather than per GPU-hour.

Find alternatives to Serverless Fine-Tuning

Metered per 1M training tokens, tiered by base-model size: under 16B $0.40; 16B-70B $2.50; 70B-300B $6.00; over 300B $10.00. The product page states users pay 'per token processed during training, not per GPU-hour', with early stopping so billing ceases when the model stops improving.

What it does

• Model training

Sources

Pricing

Pricing page · 6 Sep 2026

Dedicated 'Serverless Fine-Tuning pricing' section carrying the four model-size tiers, all four rates independently re-verified.

Sold within

Press release · 6 Sep 2026

Release describes 'Crusoe Intelligence Foundry, the managed AI platform for Crusoe Cloud' as containing Serverless Fine-Tuning; the product page repeats 'now GA in Crusoe Intelligence Foundry'.

Also from Crusoe

• Crusoe Cloud Infrastructure service

  GPU cloud running on Crusoe's own energy-colocated data centres, rented by the hour.

• Crusoe Managed Inference API service

  Hosted endpoints serving open-weight and customer-supplied models on Crusoe's own hardware, billed per token or as dedicated hourly deployments.

• Crusoe Edge Zones Infrastructure service

  Crusoe Cloud GPU capacity delivered at edge and sovereign locations via prefabricated Crusoe Spark modular units.

• Crusoe Spark Hardware

  Prefabricated modular AI data centre integrating power, cooling, fire suppression and GPU racks for deployment at a customer site.

• Crusoe Command Center Developer tool

  Unified operations and observability console for GPU fleets running AI workloads on Crusoe Cloud.

Open Crusoe in the directory

Something wrong here? Send a correction — quote this product id: crusoe-serverless-fine-tuning.