NVIDIA paper enables cross-model KV cache transfer, skipping prefill 2.7-25x faster than recompute
blaizedsouza · x · 2026-09-09
A NVIDIA research paper introduces cross-model KV cache transfer: when swapping between models in a family (routing, cascading, mid-conversation switching), the receiver reuses the source model's KV cache and skips prefill entirely, with conversion running 2.7-25x faster than reprocessing the context.
Key findings:
- LLM APIs are stateless; every turn resends the full conversation and bills it as input. Prompt caching only works on the model that produced the cache, since keys/values depend on that model's weights.
- Cross-model KV shows strong linear structure: on Qwen3 14B→32B, a single source layer explains 56% of variance in the target's keys and 32% in values, rising to 79%/65% with multiple source layers.
- The method is a closed-form ridge mapper operating per head in three steps: select top-k most predictive source layers per target layer, concatenate their KV as input, strip RoPE from keys before mapping.
Since cache hits are billed at roughly 10% of base input rate, making cache reuse portable across models could extend the biggest cost lever in LLM serving to routing and model cascades.
More from Infra
- DeepSpeed shows AutoEP/AutoTP/AutoSP parallelism gains, hosts office hours — PyTorch · 2026-10-27
- Cloudflare Durable Objects validated as top architecture for AI agent WebSocket servers — tristanbob · 2026-10-27
- Gary Marcus: With Q3 data added, Anthropic's compute-efficiency growth shows deceleration — GaryMarcus · 2026-10-27
- SpaceX locks in $40B GPU financing deal with Apollo, the largest ever committed — TechCrunch AI · 2026-10-27
- Inference is just Kubernetes with a multi-trillion-dollar market, and the communities overlap — matt_slotnick · 2026-10-27
- vLLM eyes Rust migration as inference engines race to rewrite the stack — AccBalanced · 2026-10-27