RTX 3080
4 records across 1 models (quants: Q4_0).
| hardware | model | quant | backend | tok/s | pp tok/s | W | ttft s | ctx/tg | date | class | flags | source |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| RTX 3080 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 143.44 | 4889.13 | – | – | tg128 | 2026-04-03 | community | – | source |
| RTX 3080 | llama 7B | Q4_0 | llama.cpp (CUDA) | 131.01 | 4499.47 | – | – | tg128 | 2024-12-18 | community | – | source |
| RTX 3080 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 62.16 | 1706.07 | – | – | tg128 | 2024-12-18 | community | – | source |
| RTX 3080 | llama 7B | Q4_0 | llama.cpp (Vulkan) | 18.32 | 230.79 | – | – | tg128 | 2026-04-03 | community | – | source |
All rows cite their source. Provenance: community 40018 sourced 1085