Open Models Almanac
Methodsinference & servingKV cache management

specific method · filed under inference & serving

Cross-layer KV-cache sharing

Reuses or shares KV state across model layers to reduce cache footprint.

The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See KV cache management for the methods filed beside it.

Also called Cross-layer key-value cache reuse, cross-layer KV cache reuse, Cross-layer key-value cache sharing, cross-layer KV sharing, Cross-layer KV-state sharing, shares KV state across layers.

sources
2
model
1
lab adopt it
1
strongest
core

Documented in

Mistral AI

Evidence

2 quotes from 2 sources, one card per model, each opening with its strongest.

How sources treat it
core 2
↑