Custom Rust engine runs Inkling-2-Mini at 524k context on one 24GB GPU, ~5x faster than llama.cpp

W61k3r · reddit · 2026-11-22

A developer benchmarked their custom Rust/Vulkan engine LexiEngine against llama.cpp, running Inkling-2-Mini at 524,288-token context on a single RX 7900 XTX (24GB):

Original post →

More from coding & agent

coding & agent channel →