Llama-3.2-1B

15 records across 5 hardware.

hardwaremodelquantbackendtok/spp tok/sWttft sctx/tgdateclassflagssource
GTX 1060Llama-3.2-1B Q8_0llama.cpp90.7––––2026-09-29community–source
GTX 1060Llama-3.2-1B Q8_0llama.cpp90.3––––2026-09-29community–source
IQ-9075Llama-3.2-1B Q2_Kllama.cpp (Hexagon)66.69––––2026-10community–source
IQ-9075Llama-3.2-1B Q2_Kllama.cpp (Hexagon)62.66––––2026-10community–source
IQ-9075Llama-3.2-1B Q2_Kllama.cpp (Hexagon)56.46––––2026-10community–source
IQ-9075Llama-3.2-1B Q2_Kllama.cpp (Hexagon)54.98––––2026-10community–source
IQ-9075Llama-3.2-1B Q4_0llama.cpp (Hexagon)50.33––––2026-10community–source
IQ-9075Llama-3.2-1B Q4_0llama.cpp (Hexagon)50.28––––2026-10community–source
Snapdragon 7 Gen 4 (SM7750)Llama-3.2-1B Q4_0llama.cpp (CPU)34.4216.3––tg642026-09-26community–source
Snapdragon 7 Gen 4 (SM7750)Llama-3.2-1B Q4_0llama.cpp (Hexagon)29.91749.4––tg642026-09-26community–source
Snapdragon 7 Gen 4 (SM7750)Llama-3.2-1B Q4_0llama.cpp (Hexagon)27.078.7––tg642026-09-26community–source
Adreno 840Llama-3.2-1B Q4_0llama.cpp–849––5122026-10-14sourced–source
Adreno 840Llama-3.2-1B Q4_0llama.cpp–998––5122026-10-14sourced–source
Adreno 850Llama-3.2-1B Q4_0llama.cpp–1202––5122026-10-14sourced–source
Adreno 850Llama-3.2-1B Q4_0llama.cpp–1371––5122026-10-14sourced–source

← all models