Llama 2 7B on RTX 3060
80.6best gen tok/s
2,408best prompt tok/s
640max context tested
4results
Measured · all numbers below come from the linked sources
llama.cpp
| Quant | Gen tok/s | Prompt tok/s | Source | Context tested | TTFT | Build |
|---|---|---|---|---|---|---|
| Q4_0 | 80.6 | 2,013 | llama.cpp (Vulkan) ↗ | 512+128 | — | llama.cpp Vulkan @92c0b38 — coopmat2 |
| Q4_0 | 76.9 | 2,408 | llama.cpp (CUDA) ↗ | 512+128 | — | llama.cpp CUDA @baa9255 |
| Q4_0 | 75.9 | 1,816 | llama.cpp (Vulkan) ↗ | 512+128 | — | llama.cpp Vulkan @92c0b38 — coopmat2 |
| Q4_0 | 75.6 | 2,138 | llama.cpp (CUDA) ↗ | 512+128 | — | llama.cpp CUDA @baa9255 |
Ollama
No measured Ollama results yet. Submit yours →
MLX
No measured MLX results yet. Submit yours →
"Context tested" is the largest prompt+generation length actually benchmarked (tokens), not a memory-fit estimate.