Proxima

vLLM plugin implementing the Star-KV paper

Coding & DevelopmentLLM ToolsUnknownOpen source
Visit tool

github.com

About Proxima

Proxima is a vLLM plugin that implements the Star-KV paper, resulting in faster decode kernels compared to FA2 in higher batch sizes.

Description summarised by AI from the sources listed below.

Key features

  • faster decode kernel than FA2 in higher batch sizes

Pricing

Pricing model: Unknown — we have not been able to confirm pricing from the official website, so nothing is stated here.

Pricing summarised by AI from the sources listed below.