Llama-3.2-1B
15 records across 5 hardware.
| hardware | model | quant | backend | tok/s | pp tok/s | W | ttft s | ctx/tg | date | class | flags | source |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| GTX 1060 | Llama-3.2-1B | Q8_0 | llama.cpp | 90.7 | – | – | – | – | 2026-09-29 | community | – | source |
| GTX 1060 | Llama-3.2-1B | Q8_0 | llama.cpp | 90.3 | – | – | – | – | 2026-09-29 | community | – | source |
| IQ-9075 | Llama-3.2-1B | Q2_K | llama.cpp (Hexagon) | 66.69 | – | – | – | – | 2026-10 | community | – | source |
| IQ-9075 | Llama-3.2-1B | Q2_K | llama.cpp (Hexagon) | 62.66 | – | – | – | – | 2026-10 | community | – | source |
| IQ-9075 | Llama-3.2-1B | Q2_K | llama.cpp (Hexagon) | 56.46 | – | – | – | – | 2026-10 | community | – | source |
| IQ-9075 | Llama-3.2-1B | Q2_K | llama.cpp (Hexagon) | 54.98 | – | – | – | – | 2026-10 | community | – | source |
| IQ-9075 | Llama-3.2-1B | Q4_0 | llama.cpp (Hexagon) | 50.33 | – | – | – | – | 2026-10 | community | – | source |
| IQ-9075 | Llama-3.2-1B | Q4_0 | llama.cpp (Hexagon) | 50.28 | – | – | – | – | 2026-10 | community | – | source |
| Snapdragon 7 Gen 4 (SM7750) | Llama-3.2-1B | Q4_0 | llama.cpp (CPU) | 34.4 | 216.3 | – | – | tg64 | 2026-09-26 | community | – | source |
| Snapdragon 7 Gen 4 (SM7750) | Llama-3.2-1B | Q4_0 | llama.cpp (Hexagon) | 29.9 | 1749.4 | – | – | tg64 | 2026-09-26 | community | – | source |
| Snapdragon 7 Gen 4 (SM7750) | Llama-3.2-1B | Q4_0 | llama.cpp (Hexagon) | 27.0 | 78.7 | – | – | tg64 | 2026-09-26 | community | – | source |
| Adreno 840 | Llama-3.2-1B | Q4_0 | llama.cpp | – | 849 | – | – | 512 | 2026-10-14 | sourced | – | source |
| Adreno 840 | Llama-3.2-1B | Q4_0 | llama.cpp | – | 998 | – | – | 512 | 2026-10-14 | sourced | – | source |
| Adreno 850 | Llama-3.2-1B | Q4_0 | llama.cpp | – | 1202 | – | – | 512 | 2026-10-14 | sourced | – | source |
| Adreno 850 | Llama-3.2-1B | Q4_0 | llama.cpp | – | 1371 | – | – | 512 | 2026-10-14 | sourced | – | source |