---
title: "Multi-LoRA Serving"
description: "Serving several LoRA adapters on one shared base model: what loading them together saves, and how a request for a new adapter triggers loading or eviction."
canonical_url: "https://fanout.sh/inference-eng/curriculum/26-02"
md_url: "https://fanout.sh/inference-eng/curriculum/26-02.md"
access: "pro-preview"
---

# Multi-LoRA Serving

Serving several LoRA adapters on one shared base model: what loading them together saves, and how a request for a new adapter triggers loading or eviction.

- Access: Fanout Pro (public preview)

## Public preview

Two tenants share a base model but require different adapters. Loading both adapters can save duplicating the base weights, yet a request for a third adapter may trigger loading or eviction. The service now schedules model identity as well as token work.

## In this lesson

- Adapter storage
- Cache pressure
- Engine admission
- Identity and isolation
- Comparison workload

## Continue learning

Pro unlocks the complete lesson, video, practice, and code where available.

---
This representation contains public Fanout content only. Protected Pro lessons, account data, billing, checkout, and pricing are not included.

Browse the public content map: https://fanout.sh/sitemap.md