llama.cpp adds LFM2.5-Encoder-350M and 230M model support
October 3, 2026
The llama.cpp repository now includes support for LFM2.5-Encoder models with 350M and 230M parameter sizes. This enables local inference for these specific encoder architectures within the ggml ecosystem.
HOW THIS AFFECTS YOU
●
builderYou can now run these specific small-scale encoder models locally using llama.cpp.
●
researcherThis provides a standardized implementation for testing LFM2.5 encoder performance on edge hardware.