DiggTop

LFM2.5-2.6B

Llama family · 3.0B params

2recorded runs
305.6best tok/s
2verified
Fastest GPUs for this model (best tok/s)
  1. RTX 3090305.6
  2. Intel Arc Pro B70132.4

All benchmarks Methodology

GPUModelQuantBackendCtx tok/sprompt tok/sVRAM GBSource
RTX 3090 · 24 GBLFM2.5-2.6BQ4_K_Mllama.cpp 0.3.0-dev (build 1, commit d7bd3bf)24576305.6 7609.82.3src
Intel Arc Pro B70 · 32 GBLFM2.5-2.6BQ8_0llama.cpp 9fee29e9435f; oneAPI 2026.1.18192132.4 src