Local LLM: Speeding Up LFM2.5 on a Ryzen

The same 2.6B model on the same laptop, measured across three quantizations and two backends, starting from PyTorch at bfloat16 and ending at 43.4 tokens per second.

Last updated 2026-08-05 14:16 UTC · made on claude.ai

Model, effort and cost weren't recorded for this one: it was made on claude.ai, which doesn't keep them with the page.