Lightbits Inferra Release Notes

Lightbits Inferra accelerates large-context LLM inference by offloading the KV cache to a fast storage tier, so previously-seen context is restored on demand instead of recomputed from scratch.

These notes cover one release: what it delivers, how it performs, what to know before upgrading, and the known issues open against it. Every item under Known issues describes what you may observe, how it affects you, and the recommended action while a permanent fix is in progress.

Release

Feedback

We welcome your findings from evaluating this release. Please share results, questions, or issues with Lightbits support, including the Inferra version and the model you are serving.